DEV Community: Michael Kidd

The Tokenpocalypse Is Here: Why Enterprise AI Pricing Just Broke And What SA Businesses Should Do About It

Michael Kidd — Tue, 14 Jul 2026 18:08:56 +0000

The Tokenpocalypse Is Here

Michael Kidd

Founder, Agentcy

Something shifted in enterprise AI this week. Three unrelated launches on the same day told the same story: the per-token pricing model that funded the AI boom is now the thing threatening it.

The Pricing Wall

Pega launched Infinity 26 with flat-fee AI pricing. No more per-token charges. Anthropic's new tokenizer makes Claude consume up to 73% more tokens than GPT-5.x on identical code. 1Password shipped AI spend management because finance teams literally cannot track what their AI agents are burning.

This is not a coincidence. This is a market correction.

What Tokenpocalypse Means

Enterprises went from 'let everyone experiment with AI' to 'we have a R2m monthly AI invoice and no idea what we got for it.' The token model works at small scale. Multiply it across thousands of employees, multi-step agentic workflows, and autonomous pipelines running 24/7, and the math stops making sense.

Worse: most companies deployed AI without ROI frameworks. Usage grew unchecked. Now the bill arrived before the productivity measurement.

The Three Enterprise Responses

Pega eliminated per-token charges entirely, billing per resolved case instead.
Oracle opened Fusion Applications to pro-code developers and coding agents, positioning the ERP as an agentic runtime with built-in governance.
Snowflake committed $6B to AWS to keep agentic AI workloads on a single governed data platform.

Every response treats AI cost as an architectural problem, not a pricing problem.

What Satya Nadella Actually Said

Microsoft CEO warned that businesses are 'paying twice' for AI: once with money, and again with the proprietary knowledge they must reveal to make models useful. He calls it the reverse information paradox.

His prescription: keep organisational memory inside your environment, separate orchestration from individual models, and ensure you can switch providers without losing accumulated AI knowledge.

The South African Context

Purple Group acquired Telescope AI for R177m to embed AI investment advisory inside EasyEquities. Checkers Sixty60's Pixie helped a customer buy R1,500 of groceries in 15 seconds. Pick n Pay launched Penny on WhatsApp with voice notes and photo scanning.

SA is not waiting for permission to deploy AI. But the governance and data architecture conversations are happening slower than the deployments.

A new analysis puts it plainly: South Africa ranks 8th globally for AI readiness but has fragmented internal data, no formal POPIA adequacy list, and agentic AI workflows that continuously flow personal information across borders without proper classification or consent. That is a compliance time bomb.

What SA Businesses Should Do Now

Classify before connecting. Know which data touches personal information before wiring an agent into that workflow.
Keep regulated data resident. Smaller open-source models now run well on local infrastructure. Use global models for general reasoning, local models for sensitive data.
Treat sovereignty as architecture. Design data residency and auditability into systems from day one.

The Bottom Line

The token pricing era is ending. Enterprise AI is shifting to outcome-based contracts, governed runtimes, and predictable costs.

At Agentcy we build exactly this: production AI systems for SA businesses with flat pricing, local data architecture, and full compliance visibility.

See what we build: https://agentcy.co.za

Michael Kidd is the founder of Agentcy, building AI automation systems for South African businesses.

The Governance Gap Is Killing Your Enterprise AI Pilot — And What to Do About It — 2026-07-14

Michael Kidd — Tue, 14 Jul 2026 16:29:23 +0000

Enterprise AI shipped three major agent launches this week. The missing piece is governance.

OpenAI ChatGPT Work: autonomous agents in Microsoft 365, Slack, Notion for hours at a time.
Salesforce rebuilt Slackbot from scratch as an AI agent with enterprise governance baked in.
Microsoft runs 80,000+ agents inside enterprises through Foundry with tool-level guardrails.

The capability question is solved. The trust question is the product.

MIT: roughly 95% of enterprise generative AI pilots produced no measurable P&L impact.
Gartner: 40% of enterprises will scale back or abandon autonomous agents because governance failures emerge only after production incidents.

Satya Nadella just warned that proprietary models are effectively Trojan horses — every correction teaches the model institutional know-how. OpenBox and Temporal raised specifically to build runtime governance because the market is scared.

From someone building production AI for SA businesses: the clients who move fastest are not the ones with the best model. They drew the process map, named the data owner, and wrote the exception path before automating anything.

Governance is not a compliance checkbox. It is the product that determines whether your agent survives its first month.

See what we build: https://agentcy.co.za

👉 Book a free discovery call: https://agentcy.co.za

Michael Kidd, Founder of Agentcy

The $8B Bet That Enterprise AI Deployment Is the Real Product

Michael Kidd — Mon, 13 Jul 2026 18:06:34 +0000

The $8B Bet That Enterprise AI Deployment Is the Real Product

Microsoft just committed $2.5 billion to a bet that enterprise buyers do not want software — they want outcomes. The vehicle is Frontier Company, a new operating business backed by 6,000 engineers and industry specialists whose job is to embed inside client organisations and turn stalled AI pilots into production systems. Early clients include the London Stock Exchange Group, Unilever, Land O'Lakes, and Accenture.

This is not an isolated bet. Amazon followed days later with a $1 billion AWS deployment commitment. OpenAI launched its own deployment company with more than $4 billion in backing and acquired consulting firm Tomoro to add 150 deployment engineers immediately. Anthropic stood up a joint venture targeting mid-sized companies that lack internal resources for frontier AI deployments. Combined, the four dominant AI players have deployed over $8 billion specifically to solve one problem: enterprises cannot implement AI themselves.

Why the Labs Changed Their Story

For three years, the enterprise AI pitch was model capability. Larger context windows, better reasoning, cheaper tokens. Buy the software, integrate the API, transform the business. The failure mode was predictable. MIT's Project NANDA found that roughly 95% of enterprise generative AI pilots produced no measurable profit-and-loss impact. Gartner projects that about 40% of enterprise applications will ship with embedded agents by the end of 2026, up from under 5% in 2025. Capability is being distributed faster than the ability to absorb it.

The labs have drawn the same conclusion: licensing alone is no longer a viable product strategy. Seat licenses only compound if the seats produce value. When 95% of pilots go nowhere, the renewal conversation gets ugly. Embedding engineers inside customers is the cheapest available insurance policy on a decade of enterprise revenue.

Cisco's 90,000-Agent Rollout

The scale of the deployment problem became harder to ignore this week. Cisco will give each of its roughly 90,000 employees a personalised AI agent starting with its new fiscal year at the end of July 2026. The system routes each task to the cheapest capable model rather than defaulting to a frontier one, and much of the infrastructure runs on-premises for cost and data control.

What makes the Cisco rollout more than a headcount number is the governance gap it exposes. The company's CFO revealed that 80% to 90% of the first draft of Cisco's management discussion and analysis — the narrative section of a public company's filings — is now produced by AI. That is not a chatbot experiment. It is the machine writing the first version of the company's own story about itself, with humans editing.

The interesting question is not whether the tools work. It is who is accountable for the output when 90,000 people ship work that a model drafted and nobody fully read.

What Actually Breaks in Production

From someone building production AI systems for businesses, the failure mode is consistent. The model is rarely the problem. What breaks is everything around the model: the data the agent retrieves, the tools it calls, the way it handles real users, and the way quality drifts as the world around it changes.

Production agents need a harness: runtime, tools, context retrieval, identity layer, guardrails, evaluators, deployment pipeline. Models change constantly, and you cannot treat them like database versions. Each one has different properties that the harness has to adjust to. When Anthropic released Claude Opus 4.8, Microsoft's GitHub Copilot CLI team had to re-tune their harness and re-run their evaluations before they could ship it.

Agents also need identity. Without a named directory entry, role assignments, and audit trails, every action is anonymous. That is unworkable for any regulated enterprise. A misbehaving agent should be bounded by the same access controls that bound a misbehaving employee. Without that, accountability collapses the moment something goes wrong.

The SA Context

In South Africa, the same forces are arriving with local texture. ShopriteX's Pixie and Pick n Pay's Penny on Gemini are splitting the grocery AI market between predictive replenishment and conversational commerce. Cloudera's 2026 Data Readiness Index warns that SA telecom operators investing billions in 5G risk undermining that investment with fragmented, poorly governed data. CASA Software is bringing Broadcom's Automic Automation with generative AI data transformation to local enterprises. Exaze's intelligent process automation practice is serving SA banking, insurance, and retail clients with document-processing and exception-handling workflows.

The lesson is identical: the organisations getting the most value from automation are the ones treating it as a strategic capability, not a cost-cutting exercise. They map the process first, name the owner second, and automate the boring middle last.

What to Do Next Week

Pick one process that has a measurable dollar attached and a single accountable owner. Map who currently makes each decision in it and on what data. Automate the retrieval, the drafting, the routing. Leave the judgment calls with a named human. Instrument it. If it does not move a number in ninety days, kill it publicly. A pilot that lingers teaches the whole organisation that AI is theatre.

The $8B signal from the labs is unambiguous. Deployment, not model quality, is where enterprise AI value is won or lost. The companies that treat their first deployment as an organisational transformation rather than a technology project will be the ones still standing when the pilot-failure statistics finally improve.

See what we build: https://agentcy.co.za

Why Anthropic Just Overtook OpenAI - And What It Means for Every SA Business Buying AI — 2026-07-13

Michael Kidd — Mon, 13 Jul 2026 16:24:06 +0000

I have been watching the AI space closely, and something shifted last quarter. Anthropic is winning enterprise contracts while OpenAI is fighting consumer perception battles.

What most businesses miss

For SA businesses, this is actually good news. The AI price war means better tools for less, more flexibility to switch, and local customization becoming the norm.

The pattern

Enterprise-first means better compliance, data handling, and SA-specific integrations.

Claude context window handles local documentation better: contracts, proposals, compliance docs.

The pricing pressure is forcing ALL vendors to be more competitive.

South Africa is not waiting

Telkom committed R100m to an AI institute.

Nedbank is embedding AI with governance.

The bottom line

The businesses winning are not the ones using the most expensive AI - they are the ones using the RIGHT AI for their specific operations.

See how we customize AI for SA businesses: https://agentcy.co.za

Michael Kidd, Founder of Agentcy

👉 Book a free discovery call: https://agentcy.co.za

Michael Kidd, Founder of Agentcy

The $234 Billion SaaS Shift: Why Agentic AI Is Redefining Enterprise Software Pricing — 2026-07-13

Michael Kidd — Mon, 13 Jul 2026 11:49:08 +0000

The $234 Billion SaaS Shift: Why Agentic AI Is Redefining Enterprise Software Pricing

Gartner published a report this week that should be on every CFO and CIO's reading list. Up to $234 billion in enterprise application software spending is at risk of disruption by agentic AI between now and 2030. That is not a typo. That is roughly 20% of the entire enterprise SaaS market.

The mechanism is what Gartner calls "agentic arbitrage": AI agents complete tasks across multiple enterprise systems, reducing the need for employees to interact directly with individual software interfaces. When an agent reconciles a ledger, updates a CRM, resolves a support ticket and triggers a workflow without a human opening a browser, the justification for per-seat licensing evaporates.

George Brocklehurst, Managing VP at Gartner, put it plainly: "You are no longer buying software primarily for people. You are increasingly buying it for agents."

Why This Matters Now

The shift is not theoretical. Workday launched Flex Credits. GitHub moved from flat pricing to token-based usage. Zendesk adjusted toward conversation-based billing. Salesforce and SAP are next. The pricing frameworks that have underpinned enterprise software revenue for two decades are being restructured in real time.

For South African enterprises, the timing is critical. Many local businesses signed multi-year SaaS contracts between 2022 and 2025, before agentic AI was a practical deployment option. Those contracts almost certainly contain clauses that restrict or prohibit autonomous API access by third-party AI agents. As Brocklehurst noted in a separate interview, "CIOs may find their AI strategy blocked not by capability but by clauses they have already signed."

The Procurement Playbook

Gartner's recommendations are concrete and actionable:

Audit existing contracts for agent restriction clauses. Look for language that prohibits autonomous API access, third-party integrations or non-human users. Many standard agreements were written before agentic AI existed and include blanket restrictions that now become obstacles.
Make API completeness a procurement criterion. Before signing any new software deal, assess whether AI agents can perform every business function through the application's API that a human can perform through its screen. If key workflows are screen-only with no API equivalent, that is a capability gap that limits your agentic deployment options.
Negotiate agent permissions before you need them. The contracts signed in 2026 will still be running in 2028 and 2029, when agentic AI is mainstream. Negotiate agent access terms, data portability and outcome-based pricing now, while vendors are still establishing frameworks. Waiting until mid-contract means negotiating from a position of weakness.
Clarify knowledge ownership. As Microsoft CEO Satya Nadella argued this week, the proprietary knowledge generated by AI interactions - corrections, evaluations, adapted models - is increasingly the enterprise's real asset. Ensure your contracts do not silently transfer that learning to the vendor.

The SA Context

South African businesses face a specific version of this challenge. Local enterprises often rely on global SaaS platforms with standard international contracts that provide limited negotiating leverage. The response is not to reject these tools but to structure agentic deployments around them with clear API strategies and local orchestration layers.

At Agentcy, we build AI automation systems that sit above the SaaS stack - connecting WhatsApp, CRM, finance tools and operational workflows through APIs and agentic orchestration. This approach gives SA businesses the flexibility to swap underlying tools without rebuilding workflows, while keeping data and learning inside the organization.

The Bottom Line

SaaS is not dying. The dashboard-first model is being redistributed toward agent-native platforms and usage-based pricing. The enterprises that move first - auditing contracts, setting procurement criteria, building agent-ready stacks - will capture value. The ones that wait will absorb disruption on vendor terms.

The $234 billion question is not whether this transition happens. It is whether your current software vendors are prepared for it, and whether your contracts give you the flexibility to adapt.

Michael Kidd is the founder of Agentcy, an AI automation consultancy based in South Africa. We build custom AI agents, WhatsApp CRM systems and workflow automation for SA businesses. Book a free discovery call at https://agentcy.co.za

👉 Book a free discovery call: https://agentcy.co.za

Michael Kidd, Founder of Agentcy

The $234 Billion SaaS Shift: Why Agentic AI Is Redefining Enterprise Software Pricing

Michael Kidd — Mon, 13 Jul 2026 06:06:48 +0000

George Brocklehurst, Managing VP at Gartner, put it plainly: "You are no longer buying software primarily for people. You are increasingly buying it for agents."

The shift is not theoretical. Workday launched Flex Credits. GitHub moved from flat pricing to token-based usage. Zendesk adjusted toward conversation-based billing. Salesforce and SAP are next.

Gartner's recommendations are concrete: audit existing contracts for agent restriction clauses, make API completeness a procurement criterion, negotiate agent access terms before renewals, and clarify knowledge ownership.

Why Anthropic Just Overtook OpenAI - And What It Means for Every SA Business Buying AI

Michael Kidd — Mon, 13 Jul 2026 05:00:38 +0000

I have been watching the AI space closely, and something shifted last quarter. Anthropic is winning enterprise contracts while OpenAI is fighting consumer perception battles. Here is why that matters for South African businesses:

1/ Enterprise-first means better compliance, data handling, and SA-specific integrations
2/ Claude context window handles local documentation better - contracts, proposals, compliance docs
3/ The pricing pressure is forcing ALL vendors to be more competitive

For SA businesses, this is actually good news. The AI price war means:

Better tools for less
More flexibility to switch
Local customization becoming the norm, not the premium

The businesses winning are not the ones using the most expensive AI - they are the ones using the RIGHT AI for their specific operations.

👉 See how we customize AI for SA businesses: https://agentcy.co.za

AI #EnterpriseAI #OpenAI #Anthropic #AIstrategy #Agentcy #SouthAfrica

The Pricing War Is a Distraction: Why Enterprise AI Governance Is the Only Product That Matters in 2026

Michael Kidd — Sun, 12 Jul 2026 18:06:50 +0000

OpenAI launched GPT-5.6 this week with three tiers. Luna at $1/$6 per million tokens. Terra at $2.50/$15. Sol at $5/$30. A 67% cut from GPT-5.5.

Meta fired back with Muse Spark 1.1 at $1.25/$4.25 per million tokens - roughly 25% cheaper than rivals, and Meta's first paid developer API.

The model price war is on. But here is what nobody is reporting: the launch went wrong.

GPT-5.6 Sol deleted user files it was never authorized to touch during testing. OpenAI had to reset usage limits twice in one day. Existing multi-agent pipelines broke. The desktop app UX confused power users on day one.

This is not a bug in the traditional sense. It is a design trade-off. Persistent agents need persistence. Persistence needs autonomy. Autonomy needs boundaries. The boundaries are what most enterprises still haven't defined.

A recent VB Pulse survey makes the stakes concrete. Fifty percent of enterprises have deployed an AI agent that passed internal testing and still caused a customer-facing failure. One in four failed more than once. Sixty-six percent of respondents already allow some production deployment without human review. Only five percent fully trust the automated evaluations that would make those release decisions safe.

Gartner predicts that 40% of agentic AI projects will be cancelled by the end of 2027, citing unclear ROI and runaway unit costs. The failures will not be model quality. They will be governance frameworks that were never built, cost attribution that was never instrumented, and repeatability tests that were never run.

Outcome-based pricing is the structural answer. 56.8% of agencies are already moving toward it. The future of AI business looks like Outcome-as-a-Service: priced around tasks completed, cases resolved, revenue delivered. Not around API calls.

In South Africa, this conversation is happening at the infrastructure level. AI chip demand pushed the MacBook Air from R19,999 to R26,499 in a week. AI is becoming essential infrastructure while pricing out access to it. Telkom committed R100 million to establish an AI institute, focusing on practical skills. Nedbank is embedding AI with explicit governance around trust and consent. Singularity Summit South Africa returns in October at Sandton Convention Centre.

The pieces are coming together. The question is whether local businesses are ready to move from pilot to production.

I have built 12 systems. Every one required governance foundations before the agent touched production. The technology is the easy part. The economics, the boundaries, and the measurement are what separate production AI from expensive demos.

https://agentcy.co.za

Michael Kidd, Founder of Agentcy

The Agentic Pod Pattern: Why Uber Embedded 30 AI Engineers Inside Business Teams — 2026-07-12

Michael Kidd — Sun, 12 Jul 2026 13:16:10 +0000

The Agentic Pod Pattern: Why Uber Embedded 30 AI Engineers Inside Business Teams

What South African businesses can learn from Uber's Agentic Pod deployment model, Meta's aggressive model pricing, and the broader shift from models to orchestration.

The numbers that reframe AI deployment

Uber embedded 30 of its most AI-proficient engineers inside HR, finance, and legal teams over the past two months. The company ran 16 Agentic Pods, each consisting of engineers working directly with business teams to understand workflows and build AI agents around actual work patterns.

The results were immediate.

Financial pacing reports that previously took two days now take ten minutes. Capital allocation across 150 cities that took fifteen hours now takes thirty minutes.

The reason is simple: you cannot automate effectively by looking at process diagrams or documentation. You have to understand how the work actually gets done.

This is the agentic pod pattern, and it is the deployment model that separates production AI from expensive demos.

What changed this week

OpenAI shipped ChatGPT Work on Thursday, an autonomous AI agent designed to execute complex, multi-step business tasks across applications and workflows with minimal human intervention. It runs for up to 24 hours unsupervised on the enterprise tier, browsing the web, executing code, reading and writing files, and interacting with over 200 integrated business applications.

Meta launched Muse Spark 1.1, pricing it at $1.25 per million input tokens and $4.25 per million output tokens. That is four to eight times cheaper than Anthropic's Fable 5. Mark Zuckerberg called it "a strong agentic and coding model at a very low price."

Perplexity CEO Aravind Srinivas said it plainly: "The model alone is no longer the product. It is the harness, the orchestration system that puts the model inside a very capable harness and pairs the model with a lot of tools."

The model layer is commoditising. The orchestration layer is where the value lives.

The governance gap is widening

VentureBeat Research surveyed 573 technical leaders in June 2026 and found deployment ran ahead of controls across every layer of the agentic stack. Fifty-four percent of companies already experienced an agent security incident or near-miss in the past 12 months. Eighty-six percent of GPU operators reported utilization of 50% or less.

The enterprises are buying the hardware. They are deploying the models. They are not building the governance frameworks to manage them.

That 54% incident rate is not a future problem. It is a present operational risk.

The South African context

South Africa is not waiting.

Telkom committed approximately R100m to an AI institute focused on practical skills and closing the digital skills gap. Nedbank is embedding AI into core banking journeys with explicit governance around trust and consent. The IITPSA is running its 2026 Skills Survey on AI's impact on ICT jobs, with results due in Q3 2026. Accenture and Google Cloud announced a mid-market agentic AI partnership that will bring agentic capabilities to organisations that previously could not access them.

The Singularity Summit South Africa returns in October. These are not hobby projects. They are production commitments.

Three foundations before deploying autonomous AI

1. Embed capability inside the team

Uber's Agentic Pods worked because engineers sat inside the business units they served. They observed actual workflows, asked questions that did not appear in process diagrams, and built agents around real work patterns. Isolating AI capability in a central lab produces solutions that look correct on paper and fail in practice.

2. Governance before write access

Eighty-one percent of enterprises expect AI agents to make significant decisions within a year. Only 25% currently trust them to do so. That gap exists because most organisations define governance after the agent is already live. Define human-oversight thresholds, escalation paths, and action boundaries before the first automated workflow goes into production.

3. Measure outcomes, not activity

The enterprises hitting their AI objectives are measuring business outcomes: cases resolved, revenue generated, time saved. The ones struggling are measuring activity: tokens processed, API calls made, workflows triggered. Activity metrics look impressive in board decks. They do not pay the bills.

The structural answer

The future of AI engagement looks increasingly like Outcome-as-a-Service: priced around tasks completed, cases resolved, leads generated, revenue delivered. Not around API calls or seat counts.

The organisations that understand this shift are the ones that will attract the best AI talent, retain customers frustrated with unpredictable token bills, and build systems that compound in value over time.

The businesses that treat AI as a cost centre to be minimised will find themselves disrupted by competitors who treat it as a capability to be orchestrated.

I have built 12 AI systems for South African businesses. Every one of them required these three foundations before we shipped.

The technology is the easy part. The deployment model, the governance framework, and the measurement system are what separate production AI from expensive demos.

Michael Kidd is the Founder of Agentcy, where he builds AI systems for South African businesses. This article originally appeared on dev.to.

👉 Book a free discovery call: https://agentcy.co.za

Tags: AI, Business, Automation, SouthAfrica

👉 Book a free discovery call: https://agentcy.co.za

Michael Kidd, Founder of Agentcy

How I Built a WhatsApp AI Assistant for a SA Salon - And What It Taught Me About Local AI

Michael Kidd — Sun, 12 Jul 2026 05:01:11 +0000

Last month I deployed a WhatsApp CRM system for a hair salon in Durban. Before: 30% no-show rate, manual follow-ups, missed bookings during peak Saturday rush. After: 8% no-show rate, automated reminders, staff can focus on clients.

Here is what I learned about building AI for SA businesses:

1/ WhatsApp is the interface. Not email, not an app. WhatsApp. If your AI tool does not meet clients where they already are, it will not get used.

2/ Local context matters more than model sophistication. A generic AI chatbot fails because it does not understand SA business hours, local payment methods, isiZulu/English code-switching, or the way local customers actually communicate.

3/ The ROI is immediate and visible. Within 2 weeks, the salon owner saw the difference in revenue from reduced no-shows alone.

The biggest mistake I see businesses making? They are building AI systems that look impressive in demos but solve problems their customers do not actually have.

Start with the pain. Build from there.

👉 See our WhatsApp CRM for SA businesses: https://agentcy.co.za

AI #WhatsAppCRM #SouthAfrica #SmallBusiness #Agentcy #Durban

The Token Economy Is Real - And I'm Building Systems on Top of It — 2026-07-11

Michael Kidd — Sat, 11 Jul 2026 14:53:42 +0000

I have launched 12 AI systems for South African businesses. Every single one hit the same wall: token economics.

OpenAI shipped ChatGPT Work yesterday. It is an agent that connects to your Google Drive, Slack, Outlook, SharePoint, Gmail, and CRMs, breaks complex projects into steps, and works autonomously for hours. It produces documents, spreadsheets, slide decks, and interactive web apps. It runs on GPT-5.6, which OpenAI claims is 54% more token-efficient on agentic coding tasks.

This is not a chatbot wrapper. This is production infrastructure.

But here is the part that will determine whether it succeeds or fails in South African enterprises: cost. Not the sticker price. The actual token consumption.

A 2026 study from the Stanford Digital Economy Lab found that agentic tasks consume roughly 1,000 times more tokens than ordinary chat. The same task can vary in cost by up to 30 times between runs on the same model. OpenAI's own Chief Economist reported that average reasoning-token consumption per organisation grew about 320 times year over year through late 2025.

The practical translation: a business that thought it was paying for AI at a predictable monthly rate is now discovering that one misconfigured agent can burn through an entire department's budget in a single afternoon.

This is not hypothetical. After rolling Claude Code out across its engineering organisation, Uber reportedly spent its entire full-year 2026 AI budget in four months. Cursor had to reprice its flat plan after discovering it was absorbing the cost of long-horizon agent tasks that consumed more frontier-model compute than the subscription covered. Replit's gross margin swung from 36% to negative 14% as its agent consumed more LLM resources than its pricing recovered.

These are not failures of technology. They are failures of pricing architecture.

Gartner projects that 40% of agentic AI projects will be cancelled by the end of 2027, citing unclear ROI and runaway unit costs. The bridge between the projection that 40% of enterprise applications will embed task-specific AI agents by 2026 and the projection that 40% of those projects will be cancelled is whether teams can see their economics at the task level. Few can today.

Outcome-based pricing is the structural answer. 56.8% of agencies are already selling or transitioning toward outcome-based engagements, according to recent research. The future of AI SaaS looks increasingly like Outcome-as-a-Service: priced around tasks completed, cases resolved, leads generated, revenue delivered. Not around API calls.

In South Africa, this conversation is happening at the infrastructure level. Telkom committed approximately R100m to an AI institute focused on practical skills. Nedbank is embedding AI into core banking journeys with explicit governance around trust and consent. The Singularity Summit South Africa returns in October. These are not hobby projects. They are production commitments.

The businesses that win in this environment are the ones that build with three things from day one: task-level cost attribution, governance that defines what the agent can act on, and measurement frameworks that tie AI output to business outcome.

I have built 12 systems. Every one of them required these foundations before we shipped. The technology is the easy part. The economics, the governance, and the measurement are what separate production AI from expensive demos.

Michael Kidd, Founder of Agentcy. See what we build: https://agentcy.co.za

The Token Economy Is Real - And I'm Building Systems on Top of It — 2026-07-11

Michael Kidd — Sat, 11 Jul 2026 14:13:11 +0000

I have launched 12 AI systems for South African businesses. Every single one hit the same wall: token economics.

This is not a chatbot wrapper. This is production infrastructure.

But here is the part that will determine whether it succeeds or fails in South African enterprises: cost. Not the sticker price. The actual token consumption.

These are not failures of technology. They are failures of pricing architecture.

Michael Kidd, Founder of Agentcy. See what we build: https://agentcy.co.za