The freshest development is not a model launch. It is the combination of three operating signals arriving at the same time. KPMG’s January 2026 pulse says enterprises are moving from AI experimentation toward production-grade, orchestrated agent systems; Microsoft’s 2025 Work Trend Index says 82% of leaders see this as a pivotal year to rethink strategy and operations; and Europe’s AI Act reaches a real milestone next quarter, when the law becomes fully applicable on 2 August 2026 and the Commission’s enforcement powers for GPAI obligations also kick in.
That shift is happening against a spending backdrop that makes the signal hard to ignore. Amazon reported $128.3 billion in 2025 cash capital expenditures and said those investments, primarily in technology infrastructure, are expected to increase in 2026. Meta reported $72.22 billion of 2025 capex and guided $115 billion to $135 billion for 2026. Alphabet reaffirmed roughly $75 billion of capex for 2025. Big companies are still spending like AI is infrastructure, not a side project.
What changed versus last quarter and last year is straightforward:
| Signal | Then | Now | What it means next quarter |
|---|---|---|---|
| Enterprise AI usage | 55% of organizations reported AI use in 2023 | 78% reported AI use in 2024 | The question shifts from adoption to where AI changes workflow economics first |
| Agent maturity | KPMG says reported agent deployment was 11% in Q1 2025 | 26% in Q4 2025, with leaders moving toward governed, orchestrated systems | Pilots are giving way to platform decisions, controls, and integration work |
| Governance pressure | AI Act obligations already staged in 2025 | Full applicability and GPAI enforcement powers land on 2 Aug 2026 | Legal, product, and operations teams need one operating playbook |
| Startup market structure | Broad recovery narrative | AI captures more capital while deal count stays compressed | More startups can launch, but fewer will compound without clear moats |
Theme 1 — AI strategy shifts from tool adoption to operating-model redesign
This is the biggest theme for the quarter. AI strategy is no longer mostly about picking the best chatbot, model vendor, or prompt library. As models get cheaper and closer in quality, the hard question becomes: which workflows should be redesigned end to end, and what human controls should remain in the loop? Stanford’s 2025 AI Index shows why. It reports that inference cost for GPT-3.5-level performance fell more than 280-fold between November 2022 and October 2024. It also says open-weight models narrowed some benchmark gaps from 8% to 1.7% in a year. At the same time, AI business usage rose to 78% in 2024 from 55% the year before.
That has a direct strategic consequence. When access gets cheaper and performance converges, the moat moves out of the model layer and into the operating layer. Bessemer’s 2025 State of AI makes the point well: as model performance converges, the real edge becomes knowing when, where, and why your model works, which is why evals and data lineage get more important in 2025 and 2026.
This is also where many teams will underperform next quarter. Deloitte’s year-end enterprise survey found that most organizations were still pursuing 20 or fewer experiments, and more than two-thirds said 30% or fewer of those experiments would be fully scaled in the next three to six months. That is a useful warning. The market is moving beyond curiosity, but most organizations still do not have the process discipline to scale fast.
The contrarian take is simple: cheaper AI does not make strategy easier. It makes bad strategy easier to hide. Teams can now generate more demos, more agents, and more internal excitement with less money. But unless they connect AI to a real constraint such as sales cycle length, support load, forecast accuracy, or engineering throughput, they will create activity instead of leverage. Next quarter will reward operators who redesign one important workflow, measure it tightly, and cut everything else.
CTA: Need help deciding which workflow to redesign first and how to govern it? See Advisory.
Theme 2 — Startup ops splits into two winning plays
The second theme is a split market. On one side are lean AI-native companies that can launch faster, stay smaller longer, and get more done with fewer people. On the other side are capital-hungry winners competing in markets where distribution, compute, compliance, and data advantages matter. Both plays can win. What is disappearing is the comfortable middle.
Carta’s data shows the launch side clearly. The share of new startups with a solo founder rose from 23.7% in 2019 to 36.3% in H1 2025. Carta explicitly links part of that change to AI, noting that AI has expanded what individuals can accomplish in a finite amount of time. Solo founders also hire their first employee earlier than multi-founder teams, with median time to first hire at 399 days versus 480 days.
But the funding side tells a harder story. Carta’s 2025 startup report says AI startups captured 44% of all U.S. startup capital. Its 2025 private-markets review says venture capital is concentrating into fewer, bigger, and usually AI-dominated rounds; total round count in 2025 fell to a six-year low; and at Series D, 58% of all cash raised went to AI startups. In the same report, Carta says AI valuation premiums were visible from Series A onward, reaching 193% at Series E+.
The second-order effect here matters. AI lowers the cost of starting, but it raises the cost of being ordinary. More founders can prototype, ship, and sell. Fewer founders will be able to defend margins or attention unless they own one of four things: distribution, proprietary workflow access, trusted data, or a regulated wedge that slower competitors avoid. That is why startup ops next quarter should focus less on adding more “AI features” and more on whether the product changes a business process deeply enough to be retained.
Theme 3 — Leadership becomes workforce architecture plus governance
The third theme is leadership. Not inspirational leadership. Operating leadership. Microsoft’s 2025 Work Trend Index says 82% of leaders believe this is a pivotal year to rethink core aspects of strategy and operations. It also found a clear capacity gap: 53% of leaders say productivity must increase, while 80% of the global workforce says it lacks enough time or energy to do the work. In that same report, 82% of leaders say they are confident they will use digital labor to expand workforce capacity in the next 12 to 18 months.
That sounds exciting until you look at what actually blocks progress. KPMG says 65% of leaders cite agentic-system complexity as the top barrier, 80% cite cybersecurity as the greatest barrier to achieving AI-strategy goals, 77% say data privacy is a major barrier, and 75% rank security, compliance, and auditability as the most critical requirements for agent deployment. It also says 64% of organizations have already altered their approach to entry-level hiring due to AI agents, while 76% are willing to pay up to 10% more for candidates with strong AI skills.
So the leadership job is changing in two directions at once. First, managers need to design human-agent systems: who delegates, who approves, who audits, and who handles exceptions. Microsoft calls this the “human-agent ratio.” Second, leaders need to make governance practical. Deloitte’s survey found regulation and risk became the top barrier to GenAI development and deployment, rising 10 percentage points from Q1 to Q4, even while 74% said their most advanced initiative was meeting or exceeding ROI expectations.
This is where next quarter becomes real. The EU’s AI Act is not just a policy headline anymore. The European Commission says the law becomes fully applicable on 2 August 2026, with GPAI enforcement powers also applying from that date. Alphabet’s 2025 10-K adds a broader warning: in 2025, U.S. state legislatures considered more than 1,000 AI-related bills. Even if a company is not directly in scope today, the direction of travel is clear. Documentation, transparency, testing, copyright hygiene, and incident response are moving closer to the product roadmap.
The contrarian point is that management is not disappearing. It is being redefined. The weakest managers will struggle because status updates and routine coordination are easier to automate. Strong managers will become more valuable because exception handling, prioritization, judgment, permission design, and cross-functional risk decisions matter more in an agentic environment. In other words, AI compresses clerical management and raises the premium on real management.
So what for general operators, founders, and managers
For a general audience, next quarter is not the moment to chase every new tool. It is the moment to make three hard choices. First, pick one workflow where AI can remove delay, not just create content. Second, decide what evidence you require before scaling an AI workflow across a team. Third, assign one owner for risk, auditability, and escalation before regulators or customers force the issue.
A practical 90-day checklist looks like this:
- Choose one high-friction workflow with a measurable baseline: response time, cycle time, conversion rate, QA defects, or cost per task.
- Define the human-agent split: what the system can do alone, what requires approval, and what must never be automated.
- Build a minimum evidence layer: evals, prompt and model versioning, data lineage, and an incident log.
- Audit access and permissions before deployment, especially where sensitive customer, employee, legal, or financial data is involved.
- Retrain managers first, because they will be the ones translating AI from tool usage into team behavior.
Those steps are not glamorous. That is exactly why they matter. The winners next quarter will not be the companies that talk most confidently about transformation. They will be the ones that make AI boring enough to run reliably.
Disclaimer: This article is strategic commentary, not legal, financial, or regulatory advice. If you sell into the EU, healthcare, finance, or other regulated environments, validate obligations with qualified counsel before deployment.
FAQ
What does AI strategy mean in practical terms?
In practical terms, AI strategy is not a list of tools. It is a decision about which workflows you will redesign, what results you expect, and what controls you need before scaling. That framing matters more now because AI adoption has risen sharply while costs have fallen and model performance has become less differentiated.
Why does next quarter matter more than the last one?
Because the signals are converging. Enterprises are moving from pilots to production-grade agent systems, and the EU AI Act reaches a concrete enforcement milestone on 2 August 2026. That combination makes the next quarter operational, not theoretical.[
Are AI agents mainly about cost cutting?
Not necessarily. Microsoft frames agents as “digital labor” that expands capacity, while PwC finds stronger productivity and wage outcomes in AI-exposed industries. In practice, the near-term question is often throughput, quality, and cycle time before pure headcount reduction.
What should startups prioritize first?
Startups should prioritize one painful workflow and one defensible wedge. Carta’s data shows AI is making solo and lean teams more viable, but capital is also concentrating into fewer, bigger rounds, which raises the bar for defensibility. Distribution, workflow lock-in, trusted data, and compliance depth matter more than generic AI features.
What is the biggest leadership mistake right now?
Treating AI as a side experiment owned by one team. The evidence suggests leaders now need to redesign roles, define human oversight, and operationalize governance across legal, security, product, and operations. Otherwise, pilot activity rises while business value stalls.
Do non-regulated companies need to care about AI governance yet?
Yes, because governance is increasingly becoming a customer, security, and procurement issue, not just a regulator issue. The EU timeline is tightening, and Alphabet notes that U.S. state-level AI legislation expanded sharply in 2025. Even where the law is not immediately binding, buyers and partners are already asking for evidence of control.
Leave a Reply