Chapter 13
Scenarios and No-Regret Moves
Three futures, one hedge
Scenarios and No-Regret Moves
Nobody knows exactly how far or fast this goes. It helps to plan against more than one future.
Three plausible futures
| Scenario | What it looks like | What it rewards |
|---|---|---|
| Agent-first | Most routine knowledge work happens through conversational surfaces. Seat-based software shrinks sharply. Dashboards become exception and audit views. | Moving early, rationalizing applications, building strong agent teams |
| Hybrid plateau | Agents take over a large share of support, close, procurement and similar work. Judgment-heavy and regulated work stays in applications, with assistants alongside. Two tracks run side by side. | A governed data layer below, an agentic layer above, a handful of narrow uses done well |
| Platform lock-in | A few large platforms absorb the model, the runtime, the tools, the identity and the governance. Leaving any one of them becomes very hard. | Portability, exit rights, bring-your-own-model and a control layer you own |
Many analysts treat the hybrid plateau as the base case. We'd plan for it, and hedge against lock-in.
The hedge that works in all three
In every scenario, the enterprise that owns a vendor-neutral, governed operating model is better placed:
- In an agent-first world, it's what makes agents useful instead of fluent.
- In a hybrid world, it's what keeps the two tracks consistent.
- In a lock-in world, it's the asset that stays yours when you switch platforms.
Moves that hold in every scenario
- Model the operating reality. It pays off whichever AI wins.
- Treat MCP support as a requirement. Ask every vendor how their system opens to AI, with what permissions.
- Fix identity for AI. Tie every AI action to a person's authority, and move toward distinct, scoped agent identities.
- Start narrow. A few specific jobs done well beat a broad transformation that never lands.
- Renegotiate seat-based contracts. Consumption pricing gives you leverage — and a reason to know what you actually use.
- Fund it as a multi-year transition. The operating model changes slower than the technology. Plan and staff for years, not a single budget cycle.