Model your AI future.

The Headline

Welcome to May 2028. Under this scenario the enterprise AI market index moves 79.3 to 91.6, and the center of gravity shifts toward layer 6 — agent tools. Scenario shaped by Model capability high, Cost decline high, SaaS disruption high.

What Happened to SaaS

The application layer reads strengthening (78.3 → 88.3). SaaS seat pricing renegotiated runs 6% → 22%; the question is no longer whether agents touch the workflow, but who prices the work once they do.

The Data and Context Layer

The governed-context layer reads strengthening (86 → 94.3). Enterprise data that is agent-ready moves 19% → 35%. Trusted context is what lets agents move from advice to execution.

Models, Labor, and the Open-Weight Balance

The model layer reads strengthening (87.6 → 95.7). Enterprise inference on open weights moves 23% → 32%, which sets the bargaining table for every frontier contract.

The Economics of Intelligence

Cost per governed action moves 100 → 66 (index, 100 = today). That single number is the one CFOs will quote. It reframes every model-routing decision as a financial control, not a technical preference.

Workforce and the Org Chart

Agents per human in instrumented teams moves 0.3x → 1.6x, and critical workflows that are AI-operated move 7% → 24%. The org chart is incomplete without the authority map underneath it.

Governance, Security, and Authority

Governance is the gate between insight and action. With the regulatory regime set to moderate, the durable advantage goes to the operators who can reconstruct an agent's decision path on demand.

Winners and Losers

The clearest gainer is Zeroclaw in agent tools (66 → 96). The clearest pressure lands on Microsoft in commercial tools (92 → 95). The mechanism in both cases is the same: where the work moves, the margin follows.

Opportunity Spaces and New Entrants

The clearest white space sits where growth, headroom, and fragmentation overlap — layers no incumbent has locked.

• Layer 6 — Agent Tools (opportunity 70.2/100). The frontier: cross-system agent control plane and authority graph. Watch for an agent-operations and reliability specialist.
• Layer 4 — AI-Ready Data (opportunity 59/100). The frontier: real-time retrieval plus a policy mesh. Watch for an open, hybrid retrieval-infrastructure entrant.
• Layer 5 — Developer Tools (opportunity 54.8/100). The frontier: spec-to-system agent development platform. Watch for an agent-foreman ide that keeps architecture under human control.

The Laws That Held

• Where work moves, margin follows. Layer 6 — Agent Tools gains as the work concentrates there.
• Context beats model size in bounded workflows.
• Governance must be executable, not a PDF.
• Layer 1 — AI Infrastructure shows the cost of defending the old interface against the new operator.

The Brian Letort Take

Heuristic feature vector derived directly from lever positions. Enable AI synthesis for emergent second-order effects. The operators who win 2028 are not the ones with the most software. They are the ones with the clearest authority, cleanest context, and cheapest safe intelligence. Set the levers, read the numbers, and decide what your enterprise lets agents operate.