Investor note

AI roadmap theatre: why most B2B software AI strategies will not move the growth curve

May 2026

Audience: Software investors, IC sponsors, and the chairs and CEOs they back.

The consensus inside most software boards is that an AI roadmap is an AI strategy. The list lengthens each quarter. Copilot launched. Agent in beta. Workflow assistant on the plan. Investors mark the inventory and move on. The underwriting assumption is that the roadmap will, in time, move recurring revenue, net retention, margin, or category position. The data from the first two years of disclosed AI shipping says it has not. AI roadmap theatre is the activity that signals motion to boards and markets while leaving pricing, margin, and the location of value capture untouched. A roadmap is not a strategy. Investors who do not learn to tell the difference will overpay for the next two annual cycles.

The numbers are now public. MIT’s Project NANDA found that 95% of enterprise generative AI pilots are returning no measurable financial impact against committed enterprise spend of roughly 30 to 40 billion dollars (MIT NANDA via Fortune, 2025). McKinsey’s State of AI 2025 reports 88% adoption across at least one function, with only 6% qualifying as high performers where AI contributes more than 5% of EBIT (McKinsey, 2025). The gap, McKinsey is explicit, is not technical access. It is organisational rewiring of workflows, talent, governance, and the operating model (McKinsey, 2025). The first inference for an investor is that the gap between the AI roadmap line on the earnings call and the AI line on the P&L is structural, not transitional. It will widen for any company that has not also rewired the operating model.

The second misread is on margin. ICONIQ’s January 2026 bi-annual snapshot put average AI product gross margin at 52%, up from 41% in 2024 and 45% in 2025 (ICONIQ, 2026). Bessemer’s State of AI 2025 put LLM-native company gross margin around 65%, well below the 80 to 90% ceiling that defined the prior cloud decade (Bessemer, 2025). For traditional SaaS layering AI onto an 80-dollar seat, bolt-on AI assistants add roughly 15 dollars of variable cost and drop gross margin from 80% to closer to 65% (SaaS Magazine, 2026). For every million dollars in 2026 AI product revenue, roughly 230,000 dollars walks out the door as inference cost before payroll (The SaaS CFO, 2026). The roadmap line that reads as revenue acceleration is also a margin disclosure that most investor models have not yet rebuilt for. The underwriting assumption that 80-point gross margins survive an AI roadmap is wrong on its face.

The third misread is on the moat. The dominant move on the AI line of most software 10-Ks now describes capability, not defensibility. Capability is not a moat. Morningstar’s framework is the clean one: AI tilts the moat conversation back toward data scale, switching cost engineering, and distribution into bought-and-paid-for accounts (Morningstar, 2026). HarbourVest is sharper still: three categories survive the reset with their moats largely intact, which are data moats, compliance infrastructure, and systems of record. Everything else has a clock on it (HarbourVest, 2026). Houlihan Lokey’s Q1 2026 vertical software report reaches the same conclusion from the M&A angle: vertical AI that owns proprietary workflow data is reshaping competitive moats faster than horizontal AI is doing anything visible (Houlihan Lokey, 2026). The diligence question that follows is whether the roadmap actually compounds workflow data, switching cost, or distribution. Most do not, because the underlying product was already on the wrong side of one of those three before AI arrived. AI roadmap features added to a thin moat do not thicken the moat. They make the underwriting visible.

The fourth misread is on revenue quality. ICONIQ’s report flags the structural pattern that bookings growth in 2025 and 2026 is increasingly accompanied by R&D intensity above prior norms and pricing volatility above prior norms (ICONIQ via SaaStr, 2026). Roughly 65% of SaaS vendors have layered AI consumption meters onto existing seat pricing, producing a temporary lift in average contract value that masks an unfixed business model underneath (Monetizely, 2026). Bessemer’s segmentation of AI software companies into Supernovas and Shooting Stars is a polite way of saying that two sub-cohorts now exist with materially different unit economics: roughly 25% margins with experimental pricing on one side, 60% margins after model and pricing refinement on the other (Bessemer, 2025). Multiples that treat the two as one are pricing for the average. They are mispriced at both ends.

The bull case is that the roadmap is a leading indicator and that revenue, margin, and moat impact follow with a lag of two to four quarters. The data does not support this yet. McKinsey’s 88% versus 6% gap shows the lag is operating-model lag, not deployment lag, which means it does not close with time alone (McKinsey, 2025). MIT’s GenAI Divide finding was that the cohort splitting off from the failing 95% had bought from specialised vendors and built operating partnerships at roughly twice the success rate of internal builds (MIT NANDA via Fortune, 2025). Translated for an investor, the upside is concentrated in a smaller set of names than the AI roadmap inventory implies. The downside, which is almost never modelled, is the margin compression that already shows up in the gross margin line of any company shipping AI features on legacy seat pricing.

The diligence questions are simple, and most are not being asked.

What is the workflow each AI feature changes, and is the company on the winning side of that workflow change?

Which AI line in the disclosure carries usage-based pricing that passes inference cost through, and which sits on seat pricing that absorbs it?

What proprietary workflow data, switching cost, or distribution is the roadmap compounding, and is the compounding mechanism specific to this company or available to any competitor?

What is the gross margin of the AI line stripped of cross-allocations, and what does management assume that line gets to at scale?

Of the operating-model dimensions McKinsey flagged, strategy, talent, technology, data, operating model, and adoption, which has the company actually rewired?

Watch this, not that. Watch gross margin on the AI line, the buyer changing on the AI deal, and the workflow being owned by the AI feature. Stop watching the length of the AI roadmap. The companies that move the growth curve will not be the ones with the longest roadmap. They will be the ones whose AI line, two years from now, has a margin a board can defend.

Sources

investors, moat, margin, pricing, diligence

Back to writing