DEV Community: SIKOUTRIS

Why Your Business Tools Don't Talk to Each Other (And What It's Costing You)

SIKOUTRIS — Fri, 11 Sep 2026 10:06:51 +0000

Picture this: your sales team closes a deal in the CRM, then someone manually copies the client details into the accounting software, then again into the project management tool, and again into the invoicing platform. Nobody planned for it to work this way. It just... happened.

Most growing businesses reach a point where their software stack looks less like a system and more like a collection of unrelated decisions made at different times by different people under different pressures. A CRM chosen three years ago. An accounting tool inherited from the previous bookkeeper. A project tracker the dev team added because they needed something fast.

Each tool does its job reasonably well. The problem is that none of them talk to each other.

The Real Cost Isn't the Subscription Fees

When businesses audit their software spend, they look at licence costs. Fair enough. But the deeper inefficiency is invisible on any invoice.

Think about what actually happens when data lives in silos. Someone updates a client's billing address in one place, and for weeks — sometimes months — the old address persists somewhere else. An order gets confirmed in your e-commerce platform but the fulfilment team doesn't see it until someone checks manually. A project goes over budget because the finance data and the project data never sync, so nobody caught the drift until the end of the quarter.

The friction is so routine that teams stop noticing it. It becomes the way things are done here. New employees learn the manual steps as part of onboarding, not as a workaround but as actual process.

The hidden cost isn't just hours — it's decisions made on incomplete information. Quotes sent with wrong pricing. Reports that take days to compile because the data has to be pulled from four places and reconciled by hand. Customer service reps who can't see the full picture of an account without switching between three tabs and doing mental arithmetic.

For a small or mid-sized business, this adds up fast. And unlike a bad hire or a failed campaign, it's the kind of cost that never shows up clearly enough to trigger a serious conversation.

API Integration: Not Just for Enterprises

There's a persistent myth that proper system integration — where your tools genuinely share data in real time — is something that requires an enterprise budget and a six-month IT project. It isn't, not anymore.

The shift has been gradual but significant. Modern SaaS tools are built with APIs as a default. Your CRM almost certainly has one. So does your accounting software, your e-commerce platform, your support desk. The infrastructure for connection is already there; what's missing is the layer that actually joins them together.

For straightforward workflows, platforms like Zapier or Make (formerly Integromat) handle a huge range of automations without writing a line of code. Zapier connects over 7,000 apps — that's publicly documented — and for many SMEs, the combination of a few well-designed workflows eliminates a meaningful chunk of daily manual work.

But off-the-shelf automation tools have limits. They work well for linear, predictable flows. They struggle with conditional logic, complex data transformations, legacy systems, or anything with unusual authentication requirements. When your stack includes a bespoke ERP, an industry-specific platform, or a tool that predates the modern API era, you often need something built specifically for your situation — custom API connectors built for your exact stack rather than a generic automation workflow.

The distinction matters because the goal isn't automation for its own sake. It's to get the right data to the right place without anyone having to move it manually. That's a meaningfully different brief, and it shapes what kind of solution actually solves the problem.

What to Actually Do With This

If your team is regularly copying data between systems, running manual reconciliation processes, or maintaining spreadsheets that exist purely to bridge two pieces of software — that's your diagnostic. You don't need a consultant to tell you something is broken.

Start by mapping one process: where does data enter your business, and how many times is it touched before it reaches its final destination? Most SMEs who do this exercise for the first time are genuinely surprised by the answer.

The technology to fix it almost certainly exists. The question is whether it's a workflow automation problem (fast, affordable, doable today) or a custom integration problem (more work, but a permanent fix rather than a workaround).

Either way, the status quo has a cost. It's just one that shows up in the form of hours, errors, and missed visibility — not as a line item on a software invoice.

The Hidden Cost of Making Your Team Work Around Your Software

SIKOUTRIS — Tue, 08 Sep 2026 10:12:55 +0000

The Hidden Cost of Making Your Team Work Around Your Software

Your sales rep opens three browser tabs every morning. One for the CRM, one for the quoting tool, and one for the shared spreadsheet that somehow became the source of truth for everything the other two tools can't agree on. She copies a client's address from the CRM, pastes it into the quote generator, then manually updates the spreadsheet so the operations team knows a deal is moving forward.

This takes her 12 minutes per client. She has 20 active deals. That's four hours a week, every week, that your business is paying for—not to sell, but to move data between software that was never designed to talk to each other.

Nobody decided this was acceptable. It just grew that way.

5 Signs Your Team Is Working For the Software

1. Your onboarding includes a "workaround doc."
New hires get an unofficial guide—usually a Google Doc written by someone who left—explaining how to actually use the tools. It has screenshots, warnings, and notes like "don't click that button on Fridays." If your processes require tribal knowledge to navigate, the software is the problem.

2. Spreadsheets appear wherever two tools meet.
Spreadsheets aren't bad. But when they become the connective tissue between systems—when the "real" data lives in a spreadsheet because neither system has the full picture—you're maintaining a parallel infrastructure with no audit trail and a single point of failure called Derek.

3. People preemptively "prepare" data before entering it.
When your team exports a report, reformats it in Excel, then re-imports it somewhere else, the system is forcing them to act as middleware. That's not a workflow. That's a tax.

4. Reporting requires multiple sources.
If a manager needs three exports and 45 minutes to answer "how are we doing this month?", you're not running a data-driven business—you're running a data-archaeology operation. Real decisions happen based on gut feel because the numbers are always slightly out of sync.

5. The software wins arguments.
"We can't do it that way because the system doesn't support it." If that sentence has ended a discussion in your team, the tool has become policy. Business logic—pricing rules, approval chains, exception handling—has been outsourced to a vendor who doesn't know your industry.

What Custom Business Apps Actually Change

There's a printing company in Lyon that used to manage job ticketing across four tools: an order intake form, a production scheduling app, a cost calculator, and a delivery tracker. The team exported and re-imported data twice a day. Mistakes appeared at handoff points—wrong paper stock, wrong quantities, jobs stuck in limbo.

After replacing the handoff layer with a single internal application, job errors dropped sharply in the first quarter. Not because the people changed. Because the process stopped requiring humans to act as data pipes.

That's what bespoke business applications actually solve: not the flashy stuff, but the invisible friction. The app follows your approval logic, not a generic five-step template. It speaks the language of your industry—your SKU structure, your pricing exceptions, your client tiers. It connects to what you already use without demanding you rebuild around it.

Simulators and calculators are another underrated category. A financial services consultant built a client-facing calculator that takes a prospect's situation and outputs a customized recommendation in real time. Their close rate improved—not because the product changed, but because clients arrived at discovery calls already sold on the logic.

Build vs. Buy: A Realistic Cost Frame

The honest answer is: it depends on how much your current friction costs.

Generic SaaS tools are cheap per seat but hide costs in configuration, workarounds, and things they simply won't do. Many ops teams report spending significant time monthly on manual processes that exist because the tool isn't flexible enough to automate them. That time has a price.

Custom software has a higher upfront cost, a realistic build timeline (typically 6–16 weeks for a focused internal tool), and ongoing maintenance requirements you should plan for. It's not magic and it's not always the right answer.

The break-even math usually works in favor of custom when: you have a process that runs daily and involves more than two people, the process hasn't changed fundamentally in two years (so requirements are stable), and you've already tried configuring your way out of the problem with existing tools.

If you're still in "maybe we just need a different SaaS" mode, test that hypothesis first. It's cheaper to eliminate than to custom-build.

Where custom wins decisively is when your business logic is genuinely differentiated—when the way you work is part of your competitive edge, and off-the-shelf tools flatten that edge because they were designed for the median business, not yours.

The Question That Cuts Through It

Here's a simple diagnostic: if your best-performing team member left tomorrow, how much of your operational continuity lives in their head because the software can't encode it?

If the answer makes you uncomfortable, the issue isn't talent retention. It's that your tools haven't captured how your business actually works.

That gap—between how your processes should run and how your software lets them run—is where the cost hides. It doesn't show up on an invoice. It shows up in slow deals, late deliveries, frustrated hires, and decisions made with incomplete data.

Software is supposed to enforce your rules. When it's the other way around, that's not a software problem. That's a strategy problem worth solving.

When Generic Chatbots Stop Working for Your SMB (And What to Do About It)

SIKOUTRIS — Wed, 02 Sep 2026 10:07:32 +0000

A client came to us last spring with a straightforward complaint: "Our Tidio chatbot is useless. Customers ask about their order status and it just... doesn't know anything."

They weren't wrong. They'd set it up carefully, written decent FAQ flows, and paid for the pro plan. But their business — a regional distributor with custom pricing tiers, a legacy ERP, and clients who expected account-specific information — had outgrown what a generic chatbot could offer. The tool wasn't broken. It was just the wrong tool.

This happens a lot. And the decision of whether to build a custom AI chatbot or stick with what you have is genuinely difficult. So here's what we've actually learned from going through that decision process with a dozen different SMBs.

Why off-the-shelf chatbots plateau

Tools like Intercom, Tidio, Drift, and Crisp are excellent at what they're designed for: routing conversations, handling FAQs, capturing leads, escalating to agents. They're fast to deploy, reasonably priced, and well-supported.

The wall hits when your bot needs to know things that live outside its knowledge base — things like:

A customer's current contract status
Real-time inventory levels
Pricing that varies by segment or geography
Internal ticket history from your helpdesk
Account-specific onboarding steps

Generic platforms can pull some of this via webhooks and integrations, but the further you push them, the more you're building workarounds instead of solutions. At some point you're maintaining a fragile Zapier chain just to answer "when does my invoice renew?"

The 4 situations where custom actually makes sense

1. Your chatbot needs to query your own data in real time

If accurate answers require a live database lookup — not a static FAQ — you need something built for that. A custom bot can be connected directly to your ERP, CRM, or custom database and respond with actual account information rather than generic prompts to "contact support."

2. You have a high-volume, repetitive support pattern

If your team spends a significant chunk of time answering the same 20 questions that require a data lookup (order status, availability, account balance), the ROI on a custom solution can materialize within months. We typically see break-even in 6–12 months when the automation handles 40%+ of first-contact queries.

3. Your language or domain is too specialized

Industry-specific terminology, regional language variants, technical product specs — generic bots trained on general content struggle here. A custom chatbot can be fine-tuned or prompted with your actual documentation, product catalog, and company vocabulary.

4. You need it embedded deeply in a proprietary workflow

Sometimes the chatbot isn't just a support widget — it's a step in a process. Think: a bot that helps field technicians log interventions, or one embedded in a client portal that triggers actions in your backend. That level of integration requires something built from scratch, usually as part of a broader custom application.

The 3 situations where it's probably a bad idea

1. You don't have a data problem, you have a process problem

If your support chaos comes from unclear internal processes, inconsistent information, or a team that doesn't agree on answers — a custom chatbot won't fix that. It'll just automate the confusion. Sort the source of truth first.

2. Your volume doesn't justify it

Honestly: if you're getting 30 chat requests a month, the ROI math doesn't work. Custom development (even a lightweight integration project) typically starts at several thousand euros for something maintainable. You need volume or high value-per-interaction to make it worthwhile.

3. You can't commit to maintaining it

A custom AI chatbot is not a "deploy and forget" asset. Products change. Pricing changes. Your ERP gets updated. Someone needs to own the ongoing maintenance — updating the knowledge base, monitoring edge cases, handling fallbacks when the AI gets confused. If there's no one internally who can own that, a managed off-the-shelf tool is the more pragmatic choice.

What it actually takes to build one

The technical part is genuinely more accessible than it was two years ago. LLM APIs are affordable, and frameworks for building retrieval-augmented chatbots are mature. But the real work is almost never the AI layer — it's the integrations.

In practice, a custom chatbot project breaks down like this:

Data access layer (40% of effort): connecting to your CRM, ERP, or database; handling auth; dealing with inconsistent data formats
Conversation design (25%): mapping the actual flows, defining fallbacks, writing the prompts that keep the AI on-task
UI/embedding (15%): where it lives, how it looks, mobile behavior
Testing and edge cases (20%): the AI will hallucinate sometimes; building guardrails, testing with real queries, handling graceful failures

Maintenance is ongoing. Plan for it from the start.

A word on expectations

Custom doesn't mean perfect. AI chatbots — even well-built ones — will occasionally give wrong answers, get confused by unusual phrasing, or fail gracefully when they hit a data gap. The goal isn't zero errors; it's meaningful reduction in routine workload with a system that handles edge cases better than "I don't understand, please contact support."

The clients we see happiest with their custom bots are the ones who went in with clear success metrics: deflection rate on specific query types, response time for tier-1 support, hours saved per week. Vague goals produce vague satisfaction.

If you're at that inflection point with a generic tool and wondering whether custom makes sense, we're happy to do a quick diagnostic — sometimes the answer is "not yet" and that's a useful outcome too. You can see how we approach custom chatbot development for SMBs to get a sense of the kind of projects we take on.

Either way: make the decision based on your actual data volume, your integration complexity, and your maintenance capacity — not on the promise of AI doing everything automatically. It won't. But when it's the right fit, it genuinely changes how a business operates.

5 Business Process Automation Wins for Small Businesses (Without Hiring a Developer)

SIKOUTRIS — Mon, 31 Aug 2026 10:09:25 +0000

Most small business owners don't need a developer. They need a clear answer to one question: which process should I automate first?

The tooling has genuinely matured. Zapier, Make, n8n — these platforms let a non-technical founder connect apps, trigger actions, and eliminate repetitive work without touching a line of code. But picking the right process to automate matters more than picking the right tool. Automate a broken process and you get broken results, faster.

Here are five automation wins that consistently deliver for small businesses — and what to watch out for before you flip the switch.

1. Lead Capture to CRM (The One That Pays for Everything)

Every hour a lead sits in a contact form inbox without follow-up is money evaporating. The fix is straightforward: when a form is submitted, a new contact lands automatically in your CRM, a task is created for the sales rep, and a personalised acknowledgment email goes out within 60 seconds.

McKinsey found that 30% of sales activities can be automated — and this is the obvious starting point. The ROI shows up fast because the bottleneck is visible and the fix is direct.

Watch out for: duplicate detection. If your form doesn’t check for existing contacts before creating new ones, your CRM becomes a mess within weeks.

2. Invoice Generation and Payment Follow-Up

Chasing unpaid invoices is one of the most demoralizing tasks a small business owner faces. It’s also one of the most automatable. A trigger on a completed project or closed deal can generate a PDF invoice, send it to the client, and then schedule follow-up reminders at day 7, day 14, and day 30 — all without human intervention.

This isn’t glamorous. But cash flow is the number-one killer of small businesses, and many SMEs report recovering thousands in outstanding payments simply by making the follow-up process systematic rather than guilt-driven.

Watch out for: tone. Automated reminders that sound robotic damage client relationships. Spend time writing the reminder emails once and make them feel human — that’s where the effort should go.

3. Employee Onboarding Workflows

Hiring a new person triggers a predictable cascade of tasks: IT account creation, contract signing, training materials, payroll setup, tool access. Most of this is sequential and repeatable — exactly the profile automation handles well.

A simple workflow triggered by a new hire in your HR system can send DocuSign requests, create accounts in connected apps, and assign onboarding tasks with due dates. Deloitte’s automation research consistently flags onboarding as one of the highest-value targets for HR automation in SMEs, precisely because the process is high-frequency and error-prone when done manually.

Watch out for: exception handling. New hires in unusual roles, part-time contracts, or different locations often break generic workflows. Build in a manual review step for edge cases from the start.

4. Social Media and Content Publishing Pipelines

This one divides opinion. Fully automated social media can feel hollow, and audiences notice when content is clearly scheduled-and-forgotten. But the operational side of content publishing — resizing images, scheduling posts across platforms, sending internal Slack notifications when content goes live — is pure overhead.

The smart approach is to keep the creative work human and automate the distribution mechanics. A Google Sheets or Airtable content calendar can trigger publishing to Buffer or Hootsuite, notify the team, and log the post in a master record. You get consistency without sacrificing authenticity.

5. Customer Support Triage

Not chatbots — triage. There’s an important distinction. A first-response automation that acknowledges the support ticket, assigns it to the right team member based on keywords or category, and sends the client a realistic resolution timeline is genuinely useful. It sets expectations and routes work intelligently.

Full chatbot replacement of support is a different bet entirely, and for most small businesses with complex or bespoke services, it creates more frustration than it solves. Start with triage: the automation that makes the human faster, not the one that replaces them.

What to Look for When Choosing an Automation Partner

If you’re advising SME clients — or if you’re an SME owner evaluating agencies — the key question is whether the partner builds automation around your actual processes or sells you a template. Generic automation templates rarely survive contact with a real business.

A good partner starts with a process audit: mapping what you actually do, identifying the highest-friction points, and only then recommending tooling. Agencies offering automation solutions for SMEs should be able to show you a before/after of a real workflow — not just a feature list.

McKinsey’s data puts it plainly: only 56% of automation projects deliver their projected ROI within the first year. The gap almost always comes down to poor process selection, not poor tooling.

The Right Order of Operations

Start with the process that is already working but takes too much time. Not the broken process, not the ambitious one — the boring, repeatable one that your team does correctly every single time but still hates doing.

That’s your proof of concept. Once it runs cleanly for 30 days with minimal intervention, you’ll have both the confidence and the internal buy-in to tackle something more complex. Automation compounds. Small wins early make the bigger wins possible later — and build the organizational muscle to sustain them.

Client Portals for SMEs: What Actually Works (and What Does Not)

SIKOUTRIS — Sat, 29 Aug 2026 10:35:29 +0000

Your client calls on a Tuesday afternoon. They want to know where their invoice is. You tell them to check the portal. They check the portal. The portal shows a static PDF from three months ago and a project status that was last updated manually by someone on your team who has since left. The client is frustrated. You are embarrassed. The portal, which cost real money to set up, has failed at the one job it was supposed to do.

This is not an edge case. This is the default state of most SME client portals.

The Off-the-Shelf Trap

There is no shortage of portal products aimed at small and mid-sized businesses. They come with clean dashboards, branded login screens, document libraries, and project timelines. They look impressive in demos. The problem surfaces about six weeks after go-live, when someone realizes that the portal has no idea what is happening in the CRM, cannot pull invoice status from the billing tool, and updates only when a human remembers to update it.

Off-the-shelf portals are essentially glorified file-sharing folders with a nicer UI. They are designed around a generic workflow that does not map to how any specific business actually operates. The vendors incentive is to make the demo look good. Your incentive is to make the thing work day-to-day. These two goals are not the same.

The gap shows up in three predictable places:

Billing integration: The portal does not talk to your invoicing software (Stripe, QuickBooks, Pennylane, whatever). So invoice status is always stale or missing.
Project data: Your actual project tracking lives in Jira, Notion, or a custom tool. The portal has its own project section that someone has to sync manually.
Support/comms: Clients email you anyway because the portals messaging is disconnected from your real support workflow.

What Real Integration Actually Requires

When people say "integrated portal," they usually mean a portal that can read from and write to the tools the business already uses — not a separate system that lives in isolation.

In practice, this means:

Bidirectional data flow. The portal should pull invoice status from the billing API in real time, not display a cached copy of a PDF. If a payment clears at 9pm, the portal should reflect that at 9pm — not after someone logs into two different admin panels and copies data across.

Identity that crosses system boundaries. A client logging into the portal should see their data from the CRM, their project from the PM tool, and their invoices from the billing system — all under one authentication context. This requires either a shared identity layer or a carefully managed mapping between user IDs across systems.

Webhook-driven updates, not polling. Most integrations that fail do so because they poll on a schedule. Stripe fires a webhook when an invoice is paid. Your CRM fires a webhook when a deal stage changes. Building the portal to listen to these events — rather than fetching data every hour — is the difference between a portal that feels live and one that always feels slightly out of date.

Three Patterns That Work in Practice

1. Thin frontend, thick backend. The portal UI does almost no data processing. It calls your own API layer, which aggregates data from CRM, billing, and PM tools before returning a clean response. This keeps the portal fast and maintainable. When you swap billing tools (and you will), you change one API endpoint, not the entire frontend.

2. Per-client data scoping at the database level. Row-level security in Postgres (or equivalent in whatever database you are using) means clients can only ever see their own records — enforced at the query level, not just in application logic. This is not optional. Application-level filtering fails in unpredictable ways; database-level scoping does not.

3. Event-sourced audit trail. Every meaningful action — invoice viewed, document downloaded, message sent — gets logged as an immutable event. This solves two problems: clients cant dispute what happened, and you have a debugging trail when something goes wrong. It also makes it trivial to build notification logic on top of real actions rather than polling.

A Pattern Worth Examining

A lot of the custom portal work being done for SMEs right now follows the "thin portal, rich integrations" model described above. The French digital studio Metis Digital documents their approach to custom client portals for SMEs in detail — the integration patterns they use for connecting billing, CRM, and project data, and where generic solutions tend to break down for businesses under 50 people. It is a useful reference point if you are scoping a similar project, particularly the section on authentication across multiple backend systems.

The consistent finding across projects like this: the portal itself is rarely the hard part. The hard part is the data plumbing underneath it.

The Honest Tradeoff

Custom integration work costs more upfront. A generic portal costs less on day one and more in human overhead every month after. For SMEs that have consistent, repeating client interactions — monthly reports, recurring invoices, project handoffs — the math usually favors custom within 12 months. For SMEs with highly variable, one-off client work, a well-configured generic tool with manual updates might be the right answer.

What does not work is deploying a generic portal and hoping clients will tolerate the friction. They will not. They will email you instead, and you will have paid for a portal that no one uses.

MTD ITSA Doesn't Pause for Parental Leave: What UK Freelancers Need to Know

SIKOUTRIS — Thu, 30 Jul 2026 16:11:13 +0000

August 7 is eight days away. If you're a self-employed developer, designer, or consultant with a baby on the way — or a partner who is — you're juggling two very different sets of rules at the same time. This article cuts through both of them.

Your MTD ITSA obligations don't care about your due date

Making Tax Digital for Income Tax launched this April for sole traders and landlords earning over £50,000. The first quarterly update — covering 6 April to 5 July 2026 — must be submitted to HMRC by 7 August 2026. That deadline is fixed regardless of what else is happening in your life.

There is no statutory parental leave exemption in MTD ITSA. HMRC does recognise digital exclusion as a ground for exemption, but parental leave is not on that list. What that means in practice: if you're going on leave next month, you still need to file Q1 before you go, and you'll need to file Q2 by 7 November, Q3 by 7 February 2027, and Q4 by 7 May 2027 — either yourself or through an accountant acting as your agent.

If you haven't sorted your compatible software yet, don't leave it for the weekend before. The quarterly submission deadlines are unforgiving, and choosing the right HMRC-recognised software takes longer than most people expect.

One small piece of good news: HMRC won't issue penalty points for late quarterly updates during the first year. But the Self Assessment end-of-period statement and final declaration still carry the usual consequences if missed.

Maternity Allowance: what freelancers actually get

This is where being self-employed bites hardest. Employees qualify for Statutory Maternity Pay (SMP) from their employer — up to 90% of average earnings for the first six weeks, then £184.03/week (or 90% if lower). You get none of that.

What you get instead is Maternity Allowance (MA), paid by Jobcentre Plus directly. The maximum rate is £194.32 per week (2026/27), for up to 39 weeks. You can start claiming from 26 weeks pregnant, and payments can begin from 11 weeks before your due date.

To qualify at the maximum rate, you need to have paid Class 2 National Insurance contributions for at least 13 of the 66 weeks before your due date. If you haven't been paying Class 2 — which some people on low profits choose not to — your MA could be significantly reduced, down to a minimum of around £27/week. This is one of the less-publicised traps of self-employment.

For a proper breakdown of how maternity pay in the UK works differently for the self-employed, including the exact calculation method and how to claim, the full guide covers it step by step.

One thing to plan for: MA is not paid by your client or through payroll. You apply to Jobcentre Plus, and there's a processing wait. Apply early — ideally as soon as you hit 26 weeks.

Paternity leave for the self-employed partner

If it's you supporting a partner who is giving birth, the picture is even starker. Statutory Paternity Pay (SPP) requires you to be an employee, earning at least £129/week (the Lower Earnings Limit), and to have worked for the same employer continuously for 26 weeks. Self-employed people do not qualify — full stop.

There is no equivalent of Maternity Allowance for self-employed fathers or non-birthing partners. You can take time off, of course, but you won't be paid by the state for it. Your income simply stops.

This isn't widely understood until it's too late. If you're the higher earner in a couple and you're self-employed, factor this into your finances well in advance. Some freelancers buffer this by building cash reserves, timing project work around the due date, or negotiating retainer arrangements with regular clients.

Shared Parental Leave: possible, but complicated

Shared Parental Leave (SPL) lets parents split leave between them after the first two weeks post-birth. For freelancers, the relevant scenario is usually a self-employed parent and an employed partner.

The employed partner can curtail their maternity or adoption leave and convert the remaining weeks into Shared Parental Leave — some of which the self-employed freelancer can take, though they won't receive Shared Parental Pay (ShPP) since that requires employment. The employed partner, if eligible, can receive ShPP for their own SPL weeks.

It gets complex fast. The key point: if your partner is employed and you're self-employed, it's worth checking whether SPL makes financial sense — because their ShPP weeks may be more valuable than both of you taking ad hoc time off.

A practical checklist: sort this before leave starts

Whether you're weeks or months away from a due date, there are some concrete steps worth doing now rather than later:

MTD ITSA side:

Confirm your compatible software is set up and pulling in your income/expense records correctly
Submit Q1 before you go on leave (deadline: 7 August 2026)
If you use an accountant, formally authorise them as your MTD agent so they can file on your behalf during leave
Diarise Q2–Q4 deadlines: 7 Nov 2026, 7 Feb 2027, 7 May 2027

Parental leave side:

Check your Class 2 NI contribution record — log into your Personal Tax Account on GOV.UK
Apply for Maternity Allowance as early as 26 weeks pregnant (form MA1 via gov.uk)
If your partner is employed, get their employer's SPL policy in writing
Build at least two to three months of cash buffer before leave starts — state support takes time to arrive and won't fully replace your income

The intersection of MTD ITSA and parental leave is genuinely under-documented. HMRC's guidance focuses on the tax obligations; DWP's guidance focuses on the benefits. Nobody tells you both at once.

Finance Journal covers UK tax and personal finance for freelancers and the self-employed. For more on these topics, see our maternity pay guide for the self-employed and our MTD ITSA quarterly updates guide.

MTD ITSA Late Filing Points: What Happens If You Miss the August 7 Deadline?

SIKOUTRIS — Mon, 27 Jul 2026 16:12:40 +0000

The 7 August deadline is approaching — the date by which every sole trader and landlord enrolled in Making Tax Digital for Income Tax must submit their first quarterly update for the period 6 April to 5 July 2026 (or 1 April to 30 June if you use calendar quarters). Miss it, and you might assume HMRC will immediately reach for the penalty calculator. The reality is more nuanced — and understanding exactly what happens, both now and in future tax years, could save you real money.

First, the Good News About 2026–27

For the first year of MTD ITSA — the 2026 to 2027 tax year — HMRC has confirmed there are no penalty points for missing a quarterly update deadline. This is a grace period, and it applies to all four quarterly updates for this tax year.

So if you miss the 7 August 2026 deadline, you will not receive a penalty point and you will not face a financial penalty for that missed quarterly update. HMRC has built this flexibility into the first year to allow taxpayers time to adjust to the new system.

That said, missing the update is not without consequence. You still need to send the outstanding quarterly update before you can submit your end-of-year tax return. It does not disappear — it just does not immediately cost you money. Think of it as a deadline without an immediate fine, but with an administrative burden that does not go away.

What the Points System Actually Looks Like From 2027–28

From the 2027 to 2028 tax year onwards, the grace period ends and the points-based penalty system kicks in fully for quarterly updates. This is where things get serious.

Every time you miss a quarterly update deadline, HMRC issues one penalty point. The points accumulate. Once you reach four points, a £200 financial penalty is automatically triggered. Miss another deadline after that? Another £200. And so on.

The threshold of four points is significant. With four quarterly updates per year, you could in theory hit the threshold within a single tax year — though most people are likely to miss one or two before catching up. The system is designed to penalise persistent non-filers rather than punish a one-off slip.

How Points Expire

If you accumulate points but stay below the four-point threshold, those points do not stick around indefinitely. HMRC removes them automatically 24 months after the missed deadline that caused each point.

If you reach or exceed the four-point threshold, the expiry rules change. Your points will not be removed automatically. Instead, you need to:

Submit every update and return on time for 12 consecutive months, and
Send all outstanding updates and returns from the previous 24 months

Only once both conditions are met will HMRC clear your points tally. It is a meaningful hurdle — designed to ensure the system is not gamed by filing on time for a month or two then slipping back into missed deadlines.

What About Late Payment Penalties?

The points system above relates to late submission — missing the deadline to send the quarterly update itself. Late payment of any tax owed is a separate matter, with its own penalty structure.

For payments that are up to 15 days late, there is currently no penalty. Between 16 and 30 days late, a penalty of 3% of the tax owed applies in 2026–27, rising to 4% from 2027–28. If you go beyond 30 days, penalties escalate further, and HMRC charges interest at 10% annually, calculated daily, for up to two years.

For the first year (2026–27), there is a 30-day grace period before late payment penalties apply, rather than the 15-day window that applies thereafter.

Why This Still Matters Even With the Grace Period

The lack of immediate financial penalties in 2026–27 is not an invitation to ignore the system. The quarterly update requirement does not go away — it must be filed before your tax return can be submitted. If you miss Q1 in August and Q2 in November and Q3 in February, you will face a significant backlog of outstanding submissions before you can even touch your annual return.

More importantly, building bad habits now will hurt from April 2027 onwards. The taxpayers who find themselves with four penalty points by the summer of 2028 will almost certainly be the ones who treated the first year as optional practice. Getting the process right in 2026–27 — using HMRC-compatible MTD software and establishing a routine — is far less painful than unpicking compliance failures under a live penalty regime.

For a full breakdown of what MTD ITSA requires from you — including income thresholds, who qualifies, and how quarterly updates work — the complete guide to MTD ITSA on Finance Journal is a useful starting point.

FAQ

Q: Will I get a fine if I miss the 7 August 2026 quarterly update deadline?

No — for the 2026 to 2027 tax year, HMRC has confirmed there are no penalty points or financial penalties for missing quarterly update deadlines. The grace period covers all four quarterly updates in the first year. You must still submit the update eventually, as it is required before you can file your year-end tax return.

Q: When do the penalty points start applying to quarterly updates?

From the 2027 to 2028 tax year onwards. Each missed quarterly update deadline from that point earns you one penalty point. Reach four points and a £200 penalty is issued. Each subsequent missed deadline costs another £200 on top.

Q: Can I get my penalty points removed once I have hit the four-point threshold?

Yes, but it takes sustained compliance. You need to submit all updates and returns on time for 12 consecutive months, and also clear any outstanding submissions from the previous 24 months. Both conditions must be met before HMRC will reset your points total.

German Health Insurance for Freelance Developers: GKV vs PKV Explained

SIKOUTRIS — Mon, 27 Jul 2026 13:58:07 +0000

If you are a developer relocating to Germany, or already living there and considering going freelance, health insurance will quickly become one of your biggest financial decisions. Unlike most employed positions where your employer splits the cost with you, as a self-employed person you carry the full weight yourself. Here is what you need to know.

The Two Systems: GKV and PKV

Germany runs two parallel health insurance systems:

GKV (Gesetzliche Krankenversicherung) — the statutory public system, regulated by the state
PKV (Private Krankenversicherung) — private insurance, underwritten individually by insurers

Employed workers earning below a certain income threshold are legally required to be in the GKV. Freelancers and the self-employed, however, are exempt from that obligation: you can choose either system from day one, regardless of your income.

The JAEG Threshold — And Why It Matters Less Than You Think for Freelancers

The Jahresarbeitsentgeltgrenze (JAEG), also called the Versicherungspflichtgrenze, is the income threshold above which employed workers may opt out of GKV and join PKV. For 2026 it sits at 77,400 EUR per year (6,450 EUR per month).

As a freelancer, this threshold does not gate you. You are free to sign up for PKV even if you earn 30,000 EUR a year. The choice is yours — which makes the decision both more flexible and more consequential.

What GKV Actually Costs When You Are Self-Employed

Here is the part that surprises most developers coming from employment: when you are on payroll, your employer covers roughly half the GKV contribution. When you are self-employed, you pay both halves yourself.

The general contribution rate (Allgemeiner Beitragssatz) is 14.6% of your gross income. On top of that, each statutory insurer (Krankenkasse) charges a Zusatzbeitrag (supplementary rate) — the average in 2026 is around 1.7%, though it varies by insurer from roughly 0.9% to over 2.5%.

In practice that means the full GKV burden for a self-employed person runs ~16.3% of gross income (or around 16.1% with some of the more competitive Kassen).

There is a floor: contributions are calculated on a minimum income base of 1,318.33 EUR/month in 2026, even if you earn less. And there is a ceiling: contributions are capped at the Beitragsbemessungsgrenze of 69,750 EUR/year — so no matter how much you earn above that, your GKV contribution does not increase further.

For a freelancer earning 60,000 EUR/year, the math looks roughly like this:

60,000 EUR × 16.3% ≈ 9,780 EUR/year in GKV premiums (~815 EUR/month)

That is a real cost to factor into your rate calculations.

Why Some Freelancers Choose PKV

Private health insurance can be cheaper when you are young, healthy, and earning well — and it often comes with faster specialist access, single-room hospital coverage, and no automatic family member coverage (a feature of GKV that is an advantage if you have dependants, but irrelevant and cost-neutral if you do not).

The downsides:

Premiums rise with age — sometimes steeply after 50
If your income drops or you return to employment at a lower salary, exiting PKV can be complicated
Dependants (children, non-earning partners) are not included by default, unlike GKV

Switching Back Is Harder Than You Think

One common mistake: developers join PKV early, then later want to return to GKV — perhaps because their income dropped, they had children, or they joined a startup as an employee. Getting back into GKV is possible but constrained: you generally need to be under 55 and earn below the JAEG at the new employment.

This asymmetry means the PKV decision deserves serious long-term thinking, not just a short-term premium comparison.

Where to Go From Here

The numbers above give you a framework, but the real decision depends on your specific income level, age, health history, family situation, and how long you plan to freelance. For a deeper look at the options and current costs across major providers, this detailed comparison guide for freelancers (Krankenversicherung für Selbstständige) breaks down both systems side by side in detail.

The bottom line: budget for health insurance as a real line item from the start. If you are in GKV, work backwards from ~16% of your target income to understand what daily rate you need. If you are exploring PKV, get at least three quotes and model what premiums look like at age 55, not just today.

Health insurance is not glamorous, but getting it wrong is one of the more expensive mistakes you can make as a freelancer in Germany.

UK Freelance Devs: Your First MTD ITSA Quarterly Update Is Due August 7

SIKOUTRIS — Thu, 23 Jul 2026 16:07:32 +0000

If you're a UK freelance developer, contractor, or consultant earning over £50,000 a year from self-employment, a regulatory change that came into force on 6 April 2026 affects how you report your income to HMRC. It's called Making Tax Digital for Income Tax Self Assessment — MTD ITSA for short — and your first quarterly submission deadline is 7 August 2026.

This isn't a distant future obligation. If you're in scope, Q1 has already passed (6 April to 5 July 2026), and HMRC is expecting a summary of your income and expenses for that period by 7 August.

Here's a plain-English breakdown of what it means, what you need to do, and how to actually get it done.

What Is MTD ITSA?

MTD ITSA is HMRC's programme to move Income Tax Self Assessment onto digital, real-time reporting. Instead of one annual tax return every January, you now submit quarterly updates — lightweight summaries of your income and expenses — every three months, then file a final declaration at year end.

The quarterly updates are not tax returns. They're closer to a running total: here's what I earned, here's what I spent. HMRC receives the totals only — not your individual invoices or receipts. The full end-of-year process stays largely the same.

For a deeper dive into how the whole system works, this complete MTD ITSA guide covers the sign-up process, record-keeping rules, and what the annual declaration looks like.

Who It Applies To Right Now

MTD ITSA applies if you:

Are a sole trader or landlord (or both)
Have qualifying income above £50,000 — that's total self-employment and property turnover before expenses, based on your most recent submitted Self Assessment return
Are already registered for Self Assessment

The rollout is phased:

From	Threshold
6 April 2026	Income over £50,000
6 April 2027	Income over £30,000
6 April 2028	Income over £20,000

If your day rate puts you comfortably over £50k, you're in scope now. If you're billing through a limited company and taking a small salary plus dividends, MTD ITSA does not apply — it covers sole traders and landlords, not limited company directors.

The Q1 Deadline: August 7, 2026

The MTD ITSA tax year mirrors the standard UK tax year: 6 April to 5 April. Each quarter runs as follows:

Q1: 6 April  – 5 July     → deadline: 7 August
Q2: 6 July   – 5 October  → deadline: 7 November
Q3: 6 October – 5 January → deadline: 7 February
Q4: 6 January – 5 April   → deadline: 7 May

So Q1 (6 April to 5 July 2026) has already closed, and the submission window is open now until 7 August 2026.

One useful bit of context: HMRC has confirmed it will not apply penalty points for late quarterly updates during the 2026–27 tax year. This is a soft-landing year to let people get used to the system. That said, you still need to submit before you can file your final declaration — so don't skip it entirely.

What Software Do You Need?

You cannot submit MTD ITSA updates manually or through your HMRC online account. You must use HMRC-recognised software that connects directly to their API.

The main options used by freelancers include:

FreeAgent — popular with UK contractors, has a long track record with HMRC
QuickBooks — widely used, good MTD support
Xero — strong accounting features, MTD-compliant
Sage — enterprise-leaning but has a freelancer tier
Several smaller specialist apps

HMRC's official software finder tool (on GOV.UK) lets you filter by your specific needs. If you're already using accounting software, check whether your existing provider is on the recognised list — many are.

For a side-by-side breakdown of pricing and features, this MTD-compatible software comparison covers the main options relevant to UK freelancers.

How to Actually Submit Your Q1 Update

Assuming you're signed up and using compatible software, the process is straightforward:

Make sure your records are up to date — all income received and business expenses from 6 April to 5 July 2026 should be logged in your software
Categorise your transactions — most software does this automatically from bank feeds, but review anything uncategorised
Run the quarterly update — your software will compile the totals and flag any issues
Review the summary — check the income and expense totals look right
Submit to HMRC — one click in your software; you'll get a confirmation reference

The whole thing typically takes 10–20 minutes if your records are current. The heaviest lift is the initial setup (connecting bank accounts, signing up for MTD, choosing software) — the quarterly submission itself is light.

FAQ

What if I signed up late or haven't signed up yet?

If you were supposed to be on MTD ITSA from 6 April 2026 and haven't signed up, the first step is to get compatible software and register through HMRC's sign-up service. Because 2026–27 is a penalty-free year for late quarterly submissions, you're not facing fines right now — but you do need to get set up before you can file your end-of-year declaration.

Do I submit one update per business, or one overall?

You submit separately for each source of qualifying income. If you're a sole trader with one business, that's one update. If you have self-employment income and rental income, those are submitted as separate quarterly updates. Your software should handle this automatically once you've set up each income source.

Can I submit early?

Yes. HMRC allows you to submit up to 10 days before the end of the quarter if you don't expect any further transactions in that period. You can also submit any time between the quarter end and the deadline — you don't have to wait until 7 August.

Getting This Done Before August 7

For most freelance developers, Q1 recordkeeping is probably already sitting in your accounting software — you've been invoicing clients, logging expenses, maybe reconciling your bank feed monthly. The quarterly submission is just the act of telling HMRC what the totals are.

If you're not yet set up: pick software, sign up for MTD through HMRC, import or enter your Q1 records, and submit. With a week to go before the deadline, there's still time to get it done — and with no penalties this year, the pressure is lower than it might seem.

The main thing is not to conflate "no penalties this year" with "this is optional." The quarterly updates are a legal requirement, and 2026–27 is the year to build the habit before enforcement kicks in properly.

Tour de France 2026 Stage 19: A Local's Guide to Watching the Race in Gap

SIKOUTRIS — Fri, 10 Jul 2026 03:08:56 +0000

Every summer, the Tour de France transforms quiet French towns into roaring amphitheaters of color, cowbells, and adrenaline. On July 24, 2026, it's Gap's turn — and if you've never heard of this mountain city in the Hautes-Alpes, you're about to.

Stage 19 departs from Gap and climbs toward one of cycling's most legendary finishes: Alpe d'Huez. But before the peloton tackles those famous twenty-one hairpins, it first has to get out of Gap — and that means crossing the Col de Bayard, an 8-kilometer climb averaging 5% gradient, followed by the Col du Noyer. For the riders, it's a grueling day. For spectators, it's one of the most accessible and atmospheric stages on the entire route.

I've been living in Gap for years, and I want to share what makes this place special — not just as a dot on the Tour's map, but as a genuine destination worth planning around.

Gap: A Cycling City You've Probably Never Heard Of

Gap sits at 735 meters altitude, capital of the Hautes-Alpes département, home to roughly 40,000 people. It's not Paris. It's not Nice. It doesn't have a marketing budget that puts it on Instagram travel feeds. But it has something better: 28 Tour de France passages since 1931. That's nearly a century of deep cycling heritage woven into the fabric of everyday life here.

Families in this region have been lining the cols with picnic blankets since the 1950s. It's a summer institution — grandparents who watched Louison Bobet climb these roads now sit in the same spots with their grandchildren, waiting for the helicopter noise to grow louder, signaling the peloton's approach. There's a continuity to it that you don't find in many sporting events anywhere in the world.

The Practical Stuff: How to Watch Stage 19

If you're planning to be in Gap on July 24, here's what you need to know.

The Fan Zone opens at 8:00 AM at the Plaine des Loisirs de Fontreyne, on Avenue de Traunstein, and runs until 3:00 PM. This is where the pre-race energy builds. There's a free bike repair workshop run by FDJ United and Shimano — genuinely useful if you've ridden in to watch — and free supervised bike parking at the nautical stadium, so you don't have to worry about your gear while you're cheering.

Key times:

The publicity caravan passes through Gap at 11:35 AM (bring a bag — they throw an absurd amount of freebies)
Ceremonial departure: 2:00 PM
Actual race start: 2:15 PM
Peloton hits Col de Bayard: approximately 2:23 PM

The Col de Bayard is only 11 kilometers from Gap's center, easily reachable by car or bike before closures kick in. If you want the full alpine roadside experience — folding chairs, a cooler of rosé, the distant sound of cowbells echoing off the mountains — get there before 2:00 PM.

Road Closures: Plan Ahead

This is where things get real. Road closures begin on July 23 — yes, the day before — at 4:00 PM on Avenue de Traunstein and 8:00 PM on Avenue Givaudan. If you're driving in, staying nearby, or need to navigate the city that day, plan your route in advance. The city has published detailed logistics coverage of the road closures that's worth checking before you finalize accommodation or parking plans.

More Than Just One Day

Here's something most visitors don't realize: Gap in late July is not a one-event destination. The city runs a summer festival called Éclat(s) d'Été through July and August, with free concerts at Parc Givaudan. The week of the Tour becomes a full cultural moment — cycling by day, live music in the park by evening, the particular golden light of the southern Alps stretching the sunsets past 9 PM.

The Hautes-Alpes in summer is one of France's best-kept secrets. The air is dry, the mountains are massive, and the tourist crowds that suffocate Provence or the Côte d'Azur simply don't exist here. You get Alpine grandeur without Alpine prices.

Why This Stage Matters

Stage 19 is often where Tours are decided. By this point in the race, the general classification contenders are separated by seconds. The combination of the Col de Bayard, the Col du Noyer, and the final ascent to Alpe d'Huez makes this a day where careers are defined and legends are made.

Watching it from Gap — from the departure village, from the Col de Bayard roadside, from a café terrace on the Place Jean Marcellin as the caravan rolls through — is a fundamentally different experience than watching it on television. You feel the speed. You hear the gears shift. You smell the hot tarmac and the pine trees. And for about eight seconds, the entire peloton blurs past you in a rush of Lycra and carbon fiber, and then it's gone, and you're left standing on a mountainside in the French Alps wondering what just happened.

That's the Tour. That's Gap. And on July 24, 2026, that's Stage 19.

If you've ever considered combining a cycling trip with genuine cultural immersion in a place that hasn't been overrun by tourism, Gap and the Hautes-Alpes deserve a spot on your list.

GitHub Copilot, Cursor, Codeium, or Windsurf? How I Actually Chose My AI Coding Tool in 2026

SIKOUTRIS — Thu, 02 Jul 2026 06:14:39 +0000

Picking an AI coding assistant should be simple. It is not.

A year ago I assumed I would just try the most popular one and be done with it. Two months later I had burned through four trials, read seventeen comparison posts that all said the same three things, and was still switching between tools mid-project. The problem is not that these tools are bad -- several of them are genuinely impressive. The problem is that the marketing is designed to make them all sound identical, and the actual differences only show up when you are three hours deep into a refactor.

Here is what I learned from actually using them.

The criteria that actually matter

Forget the headline benchmarks. In day-to-day use, the things that separate a good AI coding tool from a great one are:

Autocomplete quality in context. Single-line completions are basically solved. What varies wildly is how well the tool understands your file structure, your function names three files away, and the implicit conventions in your codebase. Some tools nail this; others produce confident-looking garbage.

Context window and codebase indexing. Can it actually read your whole project, or just the open file? Tools that index your repo locally versus those that send snippets to an API feel completely different in practice -- especially on larger codebases.

IDE integration depth. A plugin that sits in VS Code is not the same as a tool built around an editor. Inline diffs, multi-file edits, and chat that can reference your terminal output are increasingly table stakes, but not everyone delivers them equally.

Pricing model. The difference between per-seat monthly, usage-based, and free tiers matters enormously depending on whether you are a solo dev, a freelancer, or on a team. Some tools are genuinely free for individuals; others will surprise you with a bill after a heavy week.

Privacy and offline behavior. If you work on proprietary code, knowing what gets sent where matters. Some tools have explicit zero-retention modes; others are vague about it. Tabnine in particular has carved out a niche here.

The tools, honestly

GitHub Copilot is the safe choice -- and that is both its strength and its limitation. It works everywhere, the VS Code integration is mature, and it handles boilerplate and repetitive patterns better than almost anything else. Where it falls short is genuine reasoning: it is great at completing what you started, less good at figuring out what you should be doing. Pricing has gotten more competitive but the business tier gets expensive on larger teams. If you are already deep in the GitHub ecosystem, it is hard to argue against it as a baseline.

Cursor is what convinced me that the editor-native approach is qualitatively different. The multi-file editing, the ability to open a chat that actually knows what is in your terminal, the composer view -- these are not gimmicks. They change how you work. The downside is that it is another editor to maintain, and if you are on a team that standardizes on something else, the onboarding friction is real. It also costs more than Copilot at the individual tier. From what I have seen, it is the tool that senior engineers tend to adopt once they have tried a few others.

Codeium punches above its weight for a free product. The autocomplete is solid, it supports a genuinely large number of IDEs, and the context-awareness is better than you would expect from something with a free tier. The enterprise feature set is less developed than Copilot or Cursor, and the chat experience feels more like a bolt-on than an integrated workflow. But if budget is a constraint and you want something that works well in your existing editor without much configuration, it is the honest recommendation.

Windsurf (from Codeium's team) is the most interesting new entrant. The "flow" model -- where the AI maintains state across a longer editing session rather than responding to individual prompts -- genuinely reduces the context re-explaining overhead that gets tedious in other tools. It is still maturing, and some of the more ambitious multi-step edits produce mixed results, but the direction is right. Developers I have talked to who do a lot of greenfield work seem to find it more natural than tools optimized for completion.

Tabnine is the one I recommend to people who work on codebases with compliance requirements. The local model option, the explicit data-not-leaving-your-machine guarantees, and the enterprise security posture are genuinely differentiated. The trade-off is that the completions are weaker than the cloud-based competitors, and the reasoning capabilities are more limited. If privacy is table stakes, it belongs in your evaluation. If it is not, there are better choices.

The decision framework

If I had to give a routing guide:

Solo developer, general-purpose work, budget-conscious: Start with Codeium. Upgrade to Cursor if you find yourself wanting more.
Team on VS Code, GitHub-centric workflow: Copilot is the path of least resistance and the integration is genuinely good.
Developer who spends a lot of time on complex refactors or new features from scratch: Cursor or Windsurf. Try both; they appeal to different working styles.
Enterprise or regulated environment: Tabnine for the local model, or Copilot with a Business/Enterprise plan if you want cloud with audit controls.
Someone who switches IDEs or uses Vim/Neovim: Codeium has the best multi-editor support of the group.

If you want a detailed side-by-side feature comparison, AI Coding Compare has a breakdown that covers pricing, IDE support, and model quality across 20+ tools -- useful when you are trying to verify a specific detail before committing to a subscription.

The honest truth

No AI coding tool is going to write good software for you. What the best ones do is reduce the activation energy for the parts of coding that are friction rather than thinking -- the boilerplate, the syntax you half-remember, the test scaffolding you have written a hundred times. The tool that works best is the one that fits into how you already work, not the one with the most impressive demo.

That means the answer is probably not whichever one had the best launch blog post last month. It is the one you stopped noticing after a week because it just became part of how you work.

Spend a real week on two or three of these. Not a quick spin -- a week where you use it on actual work. The difference between them becomes obvious pretty fast once you stop comparing feature lists and start comparing how often you have to re-explain yourself to the machine.

AI Act 2026 : ce que les PME françaises doivent faire concrètement

SIKOUTRIS — Fri, 19 Jun 2026 10:37:13 +0000

Le règlement européen sur l'intelligence artificielle — l'AI Act (Règlement EU 2024/1689) — est entré en application progressive depuis août 2024. Pour les PME françaises, les premières obligations concrètes arrivent en 2026. Voici ce qu'il faut retenir.

Le calendrier d'application

L'AI Act suit un calendrier échelonné :

Août 2024 : interdictions des pratiques IA inacceptables (manipulation subliminale, score social généralisé, etc.)
Août 2025 : obligations sur les modèles d'IA à usage général (GPAI), dont les grands modèles de langage
Août 2026 : obligations pour les systèmes IA à haut risque et de nombreuses autres catégories
Août 2027 : extension aux systèmes IA utilisés dans les produits couverts par d'autres réglementations

Les systèmes IA à haut risque : ce que ça couvre

La liste des systèmes à haut risque est définie dans les Annexes II et III du règlement. Elle inclut notamment :

Les systèmes d'IA utilisés dans les RH (recrutement, évaluation des employés, promotion)
L'accès aux services essentiels (crédit, évaluation du risque d'assurance)
L'éducation et la formation professionnelle
Les infrastructures critiques
Les dispositifs médicaux

Pour une PME : si vous utilisez un logiciel de tri automatique de CVs, un outil d'évaluation du risque crédit ou un système de recommandation dans une plateforme d'e-learning, vous êtes potentiellement concernés.

Les obligations pour les utilisateurs (deployers) de systèmes à haut risque

Si votre PME utilise un système IA à haut risque développé par un tiers, vos obligations incluent :

Vérifier que le système est bien conforme avant de le déployer
Former vos équipes à la supervision humaine du système
Tenir un journal de surveillance des opérations du système
Signaler les incidents graves aux autorités compétentes
Mener des évaluations d'impact sur les droits fondamentaux dans certains cas

Les obligations de transparence pour les systèmes "à risque limité"

Même les systèmes IA qui ne sont pas à haut risque peuvent être soumis à des obligations de transparence. Si votre service utilise :

Un chatbot : vous devez informer les utilisateurs qu'ils interagissent avec une IA
Un système de génération de contenu (texte, image, audio, vidéo) : le contenu doit être marqué comme généré par IA (obligation active depuis août 2026)
Un système de reconnaissance d'émotions : information obligatoire des personnes concernées

Ce que les PME doivent faire maintenant

Cartographier vos usages IA : listez tous les systèmes IA que vous utilisez (y compris les outils SaaS qui intègrent de l'IA) et identifiez leur niveau de risque selon le règlement.

Vérifier la conformité de vos fournisseurs : les éditeurs de logiciels IA doivent fournir une documentation technique et une déclaration de conformité. Demandez-la.

Former vos équipes : les équipes qui supervisent des systèmes IA à haut risque doivent comprendre le fonctionnement et les limites du système.

Mettre en place une gouvernance IA : nommez un référent IA dans l'entreprise et documentez vos décisions liées à l'IA.

Pour suivre les évolutions réglementaires et comprendre comment les nouvelles règles s'appliquent à votre secteur, des ressources spécialisées comme reglementation-ia.fr regroupent les textes officiels, guides pratiques et actualités liées à l'AI Act et aux autres réglementations numériques européennes.

Les sanctions prévues

L'AI Act prévoit des amendes pouvant atteindre :

35 millions d'euros ou 7% du CA mondial pour les violations liées aux pratiques interdites
15 millions d'euros ou 3% du CA mondial pour les violations des obligations des systèmes à haut risque
7,5 millions d'euros ou 1,5% du CA mondial pour la fourniture d'informations incorrectes

Les PME bénéficieront de plafonds réduits, mais les sanctions restent significatives.

Conclusion

L'AI Act ne concerne pas uniquement les développeurs d'IA. Toute entreprise qui utilise des systèmes d'intelligence artificielle dans ses processus doit s'interroger sur ses obligations. 2026 est l'année charnière : mieux vaut anticiper maintenant plutôt que de courir après la conformité.

Article informatif basé sur le Règlement EU 2024/1689. La situation réglementaire évolue — consultez un juriste spécialisé pour votre situation spécifique.