10 AI PREDICTIONS FOR 2026 : What business leaders should care about

I didn’t come into 2025 trying to make predictions.

But over the year, the same tensions kept surfacing in very different conversations, boardrooms debating AI spend, executives frustrated with stalled pilots, founders shipping faster than enterprises could absorb, and markets rewarding exposure to AI whether or not value had materialized.

By the end of the year, the pattern was hard to ignore.

These are predictions for 2026, but they are grounded in decisions, data, and constraints that became visible in 2025. If you’re a business leader, this isn’t about where AI is going. It’s about where your organization may already be exposed.

1. THE AI VALUATION BUBBLE DEFLATES, NOT THE TECHNOLOGY

2025 signals

• Frothy private multiples are still showing up: Perplexity was reported in talks that would value it at $18B while it had under $100M in ARR.
• Pre-launch valuation expectations are getting normalized: Mark Zuckerberg reportedly discussed backing a new AI startup (led by Yann LeCun) at around a $3B valuation, before a product was even out.
• Meanwhile, the hyperscalers are still spending like it’s wartime (capex continues to surge, even as investors start asking “where is the profit?”). cnbc.com+1

What it means

• The market is separating “AI is real” from “every AI startup is worth venture math at infinity.”
• Models keep improving, but distribution + trust + unit economics become the filter.

Business leader impact

• Vendor risk: assume pricing changes, product pivots, or consolidation among high-valuation vendors. Do not let a fast-moving startup become a single point of failure in a core workflow unless you have a clean exit path.
• Procurement discipline matters: favor short renewal cycles, data portability, and exit flexibility.

2. THE NEXT PHASE OF VALUE CREATION IN AI COMES FROM INFRASTRUCTURE

2025 signals

• McKinsey reports 71% of respondents say their org uses gen AI regularly, but only 19% say it has increased revenue by more than 5%, and only 39% report any EBIT impact, which screams “deployment and ops are now the bottleneck.” TechCrunch
• Jason Furman highlighted that 92% of U.S. GDP growth in H1 2025 came from information processing equipment and software, meaning capex is tilting toward compute and digitization. Fortune
• Governments are now funding compute access directly, Canada announced $2B for AI compute, including a $300M AI Compute Access Fund. Fortune

What it means

• The advantage shifts from “who has the fanciest model?” to “who can run AI reliably, cheaply, and safely in production?”
• Infrastructure becomes strategy: compute supply, latency, eval pipelines, observability, and compliance.

Business leader impact

• AI advantage shifts from innovation teams to operating maturity.
• Every AI initiative should be measured by cost per task, failure rate, auditability, and cycle-time reduction, not demos.
• Treat AI infrastructure like cloud FinOps matured, unmanaged costs compound quietly.

3. THE MIDDLE TIER OF NICHE SAAS HOLLOWS OUT FAST

2025 signals

• Microsoft started including Copilot for Sales and Copilot for Service for Microsoft 365 Copilot users (no extra cost), which is exactly how “good-enough features” get bundled into platforms. Gartner
• This bundling pattern shows up across the market as platforms push AI features into core suites, compressing standalone price value.

What it means

• Mid-tier horizontal tools that sold on “nice UX + light differentiation” get squeezed from both sides:
  • Above: platforms bundle AI features into existing seats.
  • Below: micro-tools and DIY workflows get cheaper to assemble.

Business leader impact

• For procurement: run a “bundle audit” every quarter. Ask, “Which paid tools are now overlapping with platform entitlements?”
• IT leadership: score vendors by replaceability (integration depth, data gravity, switching cost).
• Expect churn, consolidation, and pricing instability in this layer.

4. A CAMBRIAN EXPLOSION OF VIBE CODED MICRO-SAAS HAPPENS, BUT MOST DIE QUIETLY

2025 signals

• Cursor’s maker (Anysphere) reportedly raised at about a $9.9B valuation with roughly $500M ARR, showing how fast AI tooling is scaling. Salesforce
• GitHub says Copilot has 20M+ users and is used by 90% of the Fortune 100, and GitHub also reports 80% of new developers adopt Copilot within their first week. Deloitte Brazil+1

What it means

• Building is no longer the constraint. Distribution, trust, and differentiation are.
• The app world gets noisier. “Small and useful” floods the market. Most of it disappears.

Business leader impact

• Expect a flood of small tools, many excellent, few durable.
• Treat micro-SaaS as reversible bets: short contracts, low switching costs, clear ownership.
• Internally, allow micro-tools, but enforce data access, identity, and accountability.

5) AI-native becomes a real product differentiator

2025 signals

• McKinsey’s 2025 global survey: 62% of organizations are experimenting with AI agents, and 23% say they are already scaling at least one agentic system in a business function. That is a shift from “chat” to “systems that act in workflows.” McKinsey & Company
• Microsoft moved Copilot Studio toward agentic-first design, shipping capabilities like autonomous agents (GA), agent flows, generative orchestration (GA), and Model Context Protocol support in 2025. This is product architecture built around AI doing work, not only answering questions. Microsoft
• Microsoft Build 2025 messaging and features put agents at the center: multi-agent orchestration, Copilot Tuning, and agent governance concepts like giving agents identities through Entra. This is the platform signaling what “AI-native” looks like in enterprise software. Microsoft

What it means

• “AI-native” will stop being marketing language and become a functional category: software that is designed for AI execution, where agents are first-class entities that can plan, trigger, act, and be governed.
• “Bolt-on AI” will increasingly feel like UI decoration, useful for quick answers, but weak at changing outcomes because it does not redesign the workflow, data access, or control layer.

Business leader impact

• Buying decision: treat “AI-native” as a workflow test, not a feature check. Require vendors to show one critical workflow end-to-end where the agent can execute steps, handle exceptions, and log actions for review.
• Risk posture: bolt-on AI increases operational risk because it tends to bypass identity, audit, and process controls. Prioritize platforms where agent actions can be governed (permissions, logging, oversight), not just generated. Microsoft+1
• Competitive reality: by 2026, the advantage will not be “we added AI.” It will be “we removed steps.” If competitors use AI-native systems to compress cycle time in sales, support, finance, or onboarding, they will run faster with the same headcount.

6. ENTERPRISE AI ADOPTION BOTTLENECKS ON CONTEXT AND SKILLS, NOT CAPABILITY

2025 signals

• UBS Evidence Lab reported only 17% of surveyed firms had implemented AI “at scale,” and 59% said they were unclear on ROI.
• Gallup found adoption is still uneven across roles, and many employees are not using AI consistently, even as leadership interest rises.

What it means

• The limitation is not model IQ. It is the organization’s ability to:
  • define workflows
  • prepare data and context
  • train people
  • govern risk
  • measure outcomes

Business leader impact

• Measure progress by workflows changed, not pilots launched.
• Fund enablement as a rollout: role-specific training, standards, and benchmarks.
• If you only buy tools, you will buy activity, not outcomes.

7. AT LEAST ONE MAJOR AI LAB FACES AN EXISTENTIAL RECKONING

2025 signals

• The Financial Times reported OpenAI had an operating loss of about $7.8B in H1 2025, highlighting how brutal compute economics can be.
• Mustafa Suleyman said staying competitive at the frontier could cost “hundreds of billions” over the next 5 to 10 years, which implies the capital requirements are becoming nation-state scale. Business Insider
• Public-market nerves show up when capex meets weaker results, Oracle’s results and capex commentary triggered sharp investor reaction. The Guardian

What it means

• Someone will mis-time the cycle: overbuild, under-monetize, or get leapfrogged.
• The labs that survive look more like infrastructure companies than “software startups.”

Business leader impact

• If you depend on a single frontier vendor: build contingency. Multi-model routing, portability, and contract protections become risk management, not engineering elegance.
• Track vendor durability: pricing trends, enterprise roadmap clarity, and signs of “growth at any cost” stress.

8. GOVERNMENTS BECOME AI INFRASTRUCTURE PLAYERS, NOT JUST REGULATORS

2025 signals

• Canada’s $2B sovereign AI compute plan makes the point clear: compute access is becoming national capacity. Fortune
• Europe is building an AI infrastructure network through EuroHPC AI Factories, expanding sites and “antennas” across countries, with large-scale coordinated investment. Digital Strategy+1
• Taiwan is also moving in this direction, with Reuters reporting plans tied to national AI compute capability and advanced GPU infrastructure.

What it means

• “Where your AI runs” becomes geopolitical, not just technical.
• National compute, domestic supply chains, and strategic partnerships shape what companies can ship.

Business leader impact

• Map your AI dependency chain: GPUs, cloud regions, model providers, and data residency.
• For regulated industries: assume “AI procurement + compliance” will start looking like cybersecurity procurement.

9. PORTFOLIO FOUNDERS EMERGE, BLURRING THE LINE BETWEEN MEDIA AND SOFTWARE

2025 signals

• IAB projects U.S. creator economy ad spend to hit $37B in 2025 (26% YoY), suggesting distribution is now a first-class business asset. TV Tech
• Substack hit 5M paid subscriptions by March 2025, showing paid audience models are scaling. Financial Times+1
• Tooling is maturing: Sacra estimates beehiiv hit about $30M annualized revenue in June 2025, reflecting growing infrastructure for “audience as an engine.” Sacra

What it means

• More founders will launch with a portfolio mindset: audience, community, products, and services, all reinforcing each other.
• Media becomes the wedge, software becomes the compounding asset.

Business leader impact

• If you are a B2B company: treat creator-led distribution as a competitive channel, not a marketing experiment. Build partnerships and measurement.
• If you are a founder: design your content pipeline to feed product discovery and retention, not just reach.

10. K-SHAPED ECONOMY INTENSIFIES: AI HOLDS UP GDP WHILE OVERALL LABOR MARKET STRUGGLES

2025 signals

• Furman’s H1 2025 framing is the clearest signal: 92% of GDP growth attributed to information processing equipment and software, which implies growth is being propped up by investment in compute and digitization. Fortune
• Meanwhile, the adoption gap inside companies remains wide (scale and ROI clarity are not keeping up).

What it means

• Capital owners and high-skill leverage roles benefit first.
• The “AI dividend” is not evenly distributed without intentional workforce strategy.

Business leader impact

• Do not wait for labor market pressure to force re-skilling. Pick 3–5 job families and redesign workflows now (what gets automated, what gets augmented, what becomes higher judgment).
• Tie AI rollout to productivity metrics and role evolution plans, not just cost cutting.

A personal note

2025 was an intense year of learning for me while researching and building The Rift. Many of these perspectives came from spending time deep in the data, the markets, and the systems behind the headlines.

I have a lot planned for 2026, and I’m excited to keep bringing a grounded, practical perspective to the table. Thank you for staying on this journey with me.