DEV Community: Miloš Radić

AI Notetakers Are Useful: The Real Value Starts After the Meeting

Miloš Radić — Tue, 16 Jun 2026 14:30:33 +0000

AI notetakers became popular because they solved a problem every team recognizes: meetings are full of useful details that are easy to lose.
That made meeting capture an obvious early use case for workplace AI. It gave teams a better way to preserve what happened without relying on rushed notes, scattered recaps, or the person with the best memory.
But once the call is recorded, summarized, and searchable, the better question is what the team does with everything that was said.

Recording a Meeting Is the Easy Part

Recording and transcription are quickly becoming the baseline expectation for AI notetakers. They still matter, especially in professional services, where a small comment during a call can change what happens next.
A client might approve a change, a blocker might surface, or a follow-up might get agreed upon before everyone moves to the next call. A searchable meeting record gives the team something to return to when those details start to blur, which usually happens faster than anyone wants to admit.
That problem extends well beyond professional services. Microsoft’s 2025 Work Trend Index on the infinite workday found that 57% of meetings are ad hoc calls without a calendar invitation. When so much work happens outside neatly scheduled meetings, the record becomes a way to recover decisions made in the flow of the day.
The product category is already moving beyond plain transcripts. Summaries and action points help, but an action point still needs somewhere to go: an owner, a project, and enough context to make sense after the call disappears into the week.
At that point, capture becomes the starting line. Once calls can be recorded, transcribed, searched, and lightly interpreted, the recap is no longer the endpoint.

The Next Gap Is Turning Conversation Into Work

The next demand is more concrete: people want meeting discussions to become tasks.
In a recent study, we surveyed 256 professional services roles to examine how they use and perceive AI agents. When asked what they wanted agents to handle, meeting capture was one of the specific tasks respondents named. But they were not only asking for a recording and transcription. They wanted discussions turned into concrete tasks.
That is helpful because the work created in a meeting usually needs a clear place to go. A blocker needs an owner. A scope change needs to sit inside the active project. A sales follow-up needs to be visible to the people who will act on it.
If those items stay inside a recap, the meeting has been captured, but the work has not moved. The real gain is a cleaner handoff from conversation to assigned work, where the important outcomes become tasks with owners, context, and a clear place in the project.

The Meeting Record Should Become Project Memory

Turning discussions into tasks is useful, but Productive’s research respondents wanted one more layer: a project knowledge base that a colleague could query if the original owner was unavailable.
That matters because a task can show what needs to be done, but not always why it exists. A teammate joining later may need to know what the client approved, what changed, what is still open, or why a decision was made. In professional services firms, that context often sits with the person who was on the call. That works until someone goes on PTO, a project changes hands, or a teammate joins halfway through delivery.
A searchable project knowledge base makes the meeting useful beyond those who attended. It turns the record from a private reference into shared project memory.

The Best AI Notetakers Will Need More Than the Meeting

A meeting record is most valuable when it’s directly linked to where the actual work happens. A call might surface a blocker, confirm a scope change, or create a follow-up, but these details lose momentum when they’re isolated from the projects they impact.
That’s why the next step in the category is not just another standalone recorder, but a meeting capture that is embedded in the systems teams already use. For teams already paying for a separate notetaker, this means fewer tools and a seamless connection between calls and the people that depend on them.
That is the context for Productive’s AI Notetaker: not another place to store meeting notes, but a way to turn discussions into tasks inside software teams already use to manage projects, docs, budgets, and people.
Cleaner summaries gave AI notetakers their initial edge. The stronger test now is whether they can turn meeting capture into work the team can actually use.

AI Agent Autonomy? Yes, but Never Without Control

Miloš Radić — Mon, 08 Jun 2026 13:58:16 +0000

AI agents are usually discussed as a leap toward independence: software that can act, decide, summarize, schedule, and move work forward without waiting for another prompt. For professional services firms, the appeal is evident. The real challenge lies in establishing trust.
In a recent study, it was examined how they use and perceive AI agents. Respondents were open to agents doing more, as long as that autonomy came with control.
In the report, control appeared less as a broad principle and more as a set of practical conditions. Teams wanted agents that make their work visible, keep outputs reviewable, allow people to step in, and make data access clear.

First, Show the Work

Before professional services teams trust an AI agent to act, they need a clear view of what is happening. Productive’s research puts it simply: people want to see what the agent is doing and why.
If an agent is preparing, suggesting, updating, or summarizing something, the person responsible for the work needs to understand what is happening before they can trust the result. Visibility gives teams something concrete to judge. Without it, they are not really trusting the agent. They are trusting the surface of the answer.

Nothing Should Take Effect Without Review

Visibility helps people understand what an agent is doing. Review gives them the chance to decide whether the output should be accepted.
Respondents described “human in the loop” less as a nice extra and more as an expected standard. In professional services, that standard carries weight: unchecked output can affect billable time, delivery risk, resourcing, or client expectations.
A review is not meant to slow the agent down. It marks the point where AI assistance becomes work a person is willing to stand behind.

There Has to Be a Way to Step In

Once an agent starts moving, teams need a way to correct its course.
The research also showed that professional services teams want the ability to pause, adjust, or undo an agent’s actions. The request is practical, not overly cautious. AI can misunderstand instructions, miss context, or hallucinate.
Teams do not need agents that wait for permission at every step. They need AI agents that can keep moving without leaving people feeling locked out of the process.

Teams Need to Know What the Agent Can See

If an AI agent is going to be useful, it needs context. That is also where trust gets more complicated.
In Productive’s research, data access was one of the clearest issues concerning control. People expressed a desire to know which information an agent could use, where that data might go, and whether the agent could surface information it shouldn't. Their concerns were specific: privacy, security, GDPR, storage, and over-permissioning.
One respondent captured the risk well: an over-permissioned agent who has been asked to summarize recent attachments could expose confidential information, such as a salary spreadsheet.
Before agents can act across projects, documents, finances, or people data, professional services teams need a clear answer to the question: what data can this agent access?

The Future Is Not Full Autonomy. It Is Supervised Autonomy

The Productive’s report points to a clear threshold for professional services teams. They may be ready to hand over more repetitive work around projects, but only when they can see what is happening, review the output, step in when needed, and understand the data being used.
A 2025 Gartner survey points in the same direction. While 75% of IT application leaders said their organizations were piloting, deploying, or had already deployed some form of AI agent, only 15% were considering, piloting, or deploying fully autonomous agents that do not require human oversight.
Productive’s AI Agents match this idea of bounded autonomy. Agents can work when people do not, but within the boundaries those people set. Teams can schedule them to act on their own or prompt them directly when needed.
Professional services teams do not want AI agents that can do everything. They want agents they can trust to handle the work that slows them down, without losing control over how it gets done.

7 Real Concerns People Have About AI

Miloš Radić — Wed, 20 May 2026 14:22:06 +0000

AI is surrounded by a strange mix of excitement and concern. On one hand, it is already part of everyday work, helping with writing, analysis, and planning. On the other hand, people still talk about hallucinations, data privacy risks, and what it might mean for their jobs.
To better understand what people are actually concerned about, the professional services platform Productive surveyed 256 agency and professional services roles to examine how they use and perceive AI agents.
The results show a clear tension. Adoption is not the issue. Most people are already using AI in some form. But at the same time, many still have real concerns about how it works and where it can go wrong. Those concerns vary, but some come up more often than others, pointing to shared patterns in how people think about AI today.

1. AI Can Be Wrong, and You Still Have to Check It

Accuracy and reliability were the most common concerns in the research. People pointed to hallucinations, failure to account for the full context, and outputs that still need to be checked before anyone can trust them. With AI already in widespread use for a while, it’s likely experience that makes people more cautious.
Even when AI is useful, people still feel they have to review everything closely, which limits how much work they are willing to hand off. In other words, the concern is not just that AI can be helpful and imperfect at the same time, but that the need to verify its output doesn’t seem to go away.

2. What Happens to Your Data Still Feels Unclear

Data privacy and security remain major concerns, especially in work involving client information, internal documents, or sensitive financial data. People want to know what happens to their data once AI is involved, including which model is processing it and who can access it. That may sound technical, but it shapes whether the tool feels safe to use at all.
That concern shows up outside the workplace, too. A 2025 Consumer Trust Survey commissioned by Relyance AI found that 82% of respondents see AI data loss-of-control as a serious personal threat, while 81% suspect companies are already using their personal data for undisclosed AI training.
There is also concern about over-permissioning, where AI tools surface or access information they should not. For example, the research quotes a respondent saying that an over-permissioned agent might accidentally share a confidential salary spreadsheet because someone asked it to “summarize recent attachments.”

3. AI Taking Action Without Approval Feels Risky

Unintended or irreversible actions are a key concern, especially when AI moves beyond generating output and starts doing things on our behalf. Interestingly, this concern is more present in some roles than others. In Productive’s research, individual contributors were especially sensitive to AI taking action without explicit approval, likely because they feel more exposed when a system acts on their behalf.
The risk isn’t limited just to incorrect output. It’s the possibility that a message is sent, a change is made, or a decision is triggered before anyone gets the chance to intervene. Once an action has been taken, the damage may be harder to contain or reverse.

4. Relying on AI Too Much Could Weaken Human Skills

Erosion of human skills is one of the more familiar AI concerns, even if it was not the loudest one in the research. People have been asking some version of the same question since generative AI entered everyday work: if AI keeps taking over more thinking, writing, and problem-solving, do people get worse at doing those things themselves?
Several respondents raised that same concern. In roles where judgment develops through practice, offloading too much cognitive work could weaken the habits people need to think problems through, express ideas clearly, and solve tasks independently.

5. AI Can Flatten Voice and Make Output Sound Generic

AI-generated content can start to feel the same, even when the output is technically correct. Many people have already noticed the familiar patterns: similar tone, similar phrasing, similar structure.
Over time, those patterns make work feel less distinctive and blur the individual voice. That becomes a bigger issue in creative and client-facing work, where sounding generic is not just a style problem. It can make the output feel less thoughtful, less personal, and easier to dismiss.

6. Job Security Questions Have Not Gone Away

Job security is still on people’s minds, even if it isn’t the first concern they raise. More than panic, there is a sense of uncertainty; questions about how roles might change, which responsibilities could lose value, and what these shifts could mean for long-term stability.
Although this concern is less vocal than others, it’s no less real. It lingers in the background of adoption, shaping how people think about where AI fits and how far they want it to go.

7. The Ethical and Environmental Impact of AI Raises Real Questions

Beyond day-to-day use, broader concerns continue to surface, particularly among senior roles. People are asking about the carbon footprint of large-scale AI, the methods behind its training, and the trade-offs that shape the outputs they receive.
There is also a more practical tension underneath it all: whether automating work that feels repetitive or low-value actually creates value, or just replaces effort without improving outcomes. That makes this less about immediate risk and more about whether the direction itself feels justified.

What These Concerns Say About AI Adoption

People are already using AI. That part is not in question. But using it and feeling comfortable with it are not the same thing.
What comes through in these concerns is not resistance, but uncertainty. AI is useful, and often impressive, but also unpredictable, hard to fully understand, and sometimes difficult to control. That mix is what makes people cautious.
For now, that tension remains unresolved. AI is becoming part of everyday work faster than people can decide what they are comfortable handing over to it. The technology is moving into the workflow, but the expectations, boundaries, and trust around it are still catching up.

AI Is Revealing the Limits of Today’s SaaS Stack

Miloš Radić — Wed, 29 Apr 2026 20:12:55 +0000

Most teams in the professional services space already use AI every day. In contained situations, it works very well for them. It can quickly summarize a document, improve a draft, or analyze a clean set of inputs. But when the task at hand requires a bit more context, and the information is spread across disconnected tools, AI just won’t return satisfactory results.
The professional services platform Productive recently surveyed 256 professionals across agency and professional services roles to understand how teams are using AI today. They found that AI is used heavily for writing and summarizing, but much less for tasks that require a broader perspective, such as planning or coordination. This aligns with a familiar AI constraint: the technology only works as well as the data it can access. And when work is split across multiple systems, AI has a harder time getting the information it needs to be genuinely useful.

AI Works Well in Controlled Tasks, Not Across Workflows

AI tends to perform best when the task is clearly defined and the inputs are easy to follow. That is why it has quickly become a reliable tool for writing, summarizing, and basic analysis. The boundaries are clear, the context is contained, and the output is easy to check.
Trouble begins once the work moves beyond a single task. Real workflows are not contained in one place. They span multiple tools, evolve over time, and depend on a mix of data, decisions, and human input. In those conditions, AI is no longer working from a complete picture but piecing together fragments.
Productive’s research data on how teams actually use AI supports this. heavy use in content and summarization, but far less in planning or coordination. It’s not because these are less valuable or less popular tasks; on the contrary. They are simply harder for AI to operate in because they rely on context that spans systems rather than residing in one place.
This creates a clear ceiling for what AI can do for teams. AI can speed up parts of the work, but it cannot easily connect them. A summary might be accurate, but it does not reflect what changed in the project plan. A draft might sound right, but it does not account for the latest client decision. The more the work depends on these moving pieces, the more AI lags behind the actual state of the workflow.

Today’s SaaS Stack Splits Work Across Systems

The way most companies manage work today did not emerge by accident. Over time, SaaS tools were built to solve specific problems: one system for sales, another for project delivery, another for time tracking, another for finance, and a handful more for communication and collaboration. Each tool does its job well, but each one only sees part of the work.
A typical setup makes this clear. Deals live in a CRM, delivery happens in a project tool, hours are tracked elsewhere, financial data sits in accounting software, and key decisions are scattered across chat threads and calendars. Taken individually, these systems are effective. Together, they do not form a complete view of what is actually happening.
That fragmentation is not a failure of any single tool. It is a consequence of how SaaS evolved: products optimized for specific functions, connected through integrations, but not designed to operate as one continuous system. The result is a working environment where information is constantly moving, but rarely fully connected. For people, that means constant context switching and reconstruction; for AI, it means operating with incomplete information.
This is where the friction becomes visible. Teams spend time rechecking outputs, cross-referencing tools, and manually filling in missing pieces. What looks like a small gap at the system level becomes repeated microtasks in day-to-day work. AI does not remove that effort. In many cases, it even amplifies it.

The Direction SaaS Is Starting to Move In

There are early signs of a different pattern emerging. Instead of adding more tools or layering AI on top of existing ones, some platforms are starting to rethink how work data is structured in the first place. The focus is moving from isolated functionality toward a more connected context, where projects, financials, communication, and operations begin to live closer together.
This shift is becoming more relevant as building software gets easier. With AI and rapid development tools lowering the barrier to launching new apps, features are faster to replicate than ever. New tools can be built, copied, and iterated on quickly. What is harder to replicate is the underlying context: how well a system brings together the data that reflects how work actually happens.
Some platforms are already moving in this direction. Productive is one example, built from the start to bring project, financial, and operational data into one place, and now introducing AI capabilities such as AI agents and AI Notetaker to operate within that context.
When there’s a single work context, AI stops being limited to isolated tasks. It can follow how work moves, pick up on changes, and act on information that reflects the full state of a project, not just a single input.