Operator note

How to tell whether your AI strategy is theatre

May 2026.6 min read

Audience: CEOs and senior operators inside B2B software companies running an AI programme.

Most software companies have an AI strategy on paper. A small share have one in their operating model. The gap between the two has a name. AI roadmap theatre. The roadmap is real, the slides are clean, the demos work, the press release ran. None of it is moving the lines on the P&L that should be moving. The diagnostic below is built for an operator who is suspicious of their own narrative. Ten minutes. Five tells. Three or more, and the AI strategy is theatre.

The operating tension

CEOs and operators face an asymmetric incentive on AI. The cost of being seen as slow is high and immediate. The cost of running a programme that does not reset the economics is hidden and lagging. So the default is motion: ship the feature, brief the board, post the case study, attend the conference. Motion is cheap. Reset is expensive. The two do not look different until the renewal cycle, the gross margin disclosure, or the analyst question forces the difference into the open.

The false diagnosis

When a CEO senses the gap, the usual first move is to demand more output. More features, faster, with more visibility. That accelerates the theatre, because output was never the bottleneck. The bottleneck is that the operating model has not changed. The five tells below isolate that bottleneck.

Tell one. The roadmap is feature-led, not workflow-led

A feature-led roadmap reads as a list of capabilities added to existing screens. A workflow-led roadmap reads as a re-architecture of what work the product owns end-to-end and what it now resolves rather than supports. The honest test is to take the roadmap to a customer success team and ask which slides correspond to a workflow the product can now complete without a human in the middle. If the answer is none, the roadmap is feature-led. That is the first tell.

The wedge is sharp. A copilot is not a moat. A copilot or AI feature, however strong in a demo, is not defensible if it does not own the workflow, the distribution, or proprietary data (Bessemer Venture Partners, 2026). Model capability is not a competitive moat. A feature-led AI roadmap is a list of copilots.

Tell two. Pricing has not changed

Pricing is the most direct test that the AI strategy is real. If the product has added AI capability and the price list has not moved, the company is giving the capability away inside a seat licence and absorbing the inference cost on its own gross margin.

That is not strategy. That is subsidy.

The market data is unambiguous. Per-seat pricing collapsed from 21 percent of SaaS to 15 in the twelve months to early 2026 (Monetizely, 2026). Eighty-three percent of AI-native SaaS now bills on usage, and Zendesk and Intercom moved to outcome-based pricing for AI-resolved work (Stormy AI, 2026). Gartner forecasts 40 percent of enterprise SaaS will move to outcome billing by 2030. If the pricing model is unchanged, the company is the laggard in its own category.

Tell three. Gross margin assumptions have not been revisited

The 80-point gross margin that defined SaaS was the consequence of multi-tenant cost-to-serve at near-zero variable cost. Inference is not free. ICONIQ data puts the average AI product gross margin at 52 percent, and inference at 23 percent of revenue at scaling-stage AI B2B companies (Monetizely, 2026). Public SaaS companies disclosing AI-driven margin pressure now report gross margins 10 to 17 points below pre-AI baselines, and inference-cost ratios in the 4 to 9 percent of revenue range called out in MD&A (SaaS Mag, 2026). If the board pack is still anchored to 80, the AI strategy has not yet reached the finance line.

That is the AI margin illusion. AI revenue that reads as growth but carries inference cost, support burden, and implementation labour that erode the margin it appears to create. The illusion can survive one or two reporting cycles. It does not survive a renewal cycle.

Tell four. The go-to-market motion is unchanged

A real AI strategy reaches the sales motion. The buying committee, the discovery questions, the proof points, the renewal model. If the rep is still leading with seat counts and the customer success team is still measuring adoption by login frequency, the go-to-market motion is calibrated to the first growth curve. Median committees for deals above 50,000 dollars now sit at 11.2 stakeholders, up from 9.7 in 2024, with two to four weeks added to enterprise cycles for procurement and AI governance (Corporate Visions, 2026; Martal, 2026). A motion designed for a 30-to-60-day, three-stakeholder cycle is not what the market is buying.

Tell five. The operating model and capital allocation have not shifted

The honest test is on the spreadsheet, not the slide. Where is the marginal dollar of operating expense going. If AI is being funded out of slack inside engineering and marketing, the company is running an AI programme. If the operating model has been redesigned, with an inference budget owned by finance, an AI-adjusted gross margin tracked monthly, and a concentrated capital bet behind one workflow-collapse opportunity, the company is running an AI strategy. McKinsey reports 5.8x average ROI on production AI within 14 months, but only 29 percent of executives see significant returns, and 79 percent of organisations report adoption challenges (McKinsey, 2025; Writer, 2026). The 50-point gap between average and median is the cost of running AI without redesigning the operating model.

Why the intervention is sequencing, not effort

The theatre survives when the company tries to fix the symptoms. More features, more demos, more press. The intervention is sequencing. Reset pricing first. Recut gross margin assumptions second. Redesign the go-to-market motion third. Concentrate capital on one bet fourth. Reprice the roadmap fifth.

In that order, each step makes the next one easier. In the reverse order, each step makes the next one harder. Most theatre programmes are in the reverse order.

The leadership implication

The CEO question that flushes the theatre is direct. Which line of the P&L has the AI strategy moved in the last twelve months, and which line will it move in the next twelve. If neither answer is specific, the company has an AI roadmap, not an AI strategy.

The rule of thumb

Run the five tells once a quarter. Score honestly. If three or more are true, suspend the next feature commitment until the operating model has caught up. The cost of running an AI strategy as theatre is paid at renewal. The cost of pausing one feature is paid in a sprint. Pay the smaller bill.

Sources

CEO, AI strategy, diagnostic, pricing, gross-margin

Back to writing