The Writer 2026 enterprise AI survey landed in April with a finding most boardrooms would prefer to ignore. Three quarters of C-suite executives admit their company’s AI strategy is “more for show” than genuine guidance, more than half say AI adoption is tearing their organisation apart, and only 29% see significant return on investment despite individual productivity gains reported at five times pre-AI baselines. The sample was 1,200 executives and 1,200 non-technical employees actively using AI at work, large enough to make the result hard to dismiss as anecdote or selection bias.
That is the headline. The unsettling part sits underneath. The same C-suite that calls its own strategy theatrical is also planning layoffs for non-adopters, cultivating an internal “AI elite”, and pouring more than a million dollars a year into platforms it has no documented plan to monetise. Investment is climbing. Conviction is not. The gap between what leaders publish and what they actually believe has become a strategic problem in its own right.
For independent CRM consultants the pattern is familiar. The same arc played out with Salesforce in 2014, with marketing automation in 2018, and with low-code platforms in 2022. The tools were genuine. The strategy was not. What is different this time is the speed of the gap, and the depth of the cultural damage when leadership mistakes activity for direction. The cost of theatre is no longer just a missed quarter. It is the trust of the operating team.
Why does the C-suite call AI strategy ‘for show’?
Seventy-five percent of executives in the Writer 2026 survey said their company’s AI strategy was published more to satisfy boards, analysts, and investors than to guide product, operations, or revenue decisions. The clearest tell is that 39% have no formal plan linking AI tools to revenue lift, despite spending above a million dollars a year on those tools. Decks reference autonomy, agents, and orchestration. Operating models still budget on headcount per process and reward managers for the size of their teams. When the words and the incentives disagree, the words lose.
Theatre is not laziness. It is what happens when a leadership team feels obligated to publish a position before it has done the discovery work. Generative AI made that pressure ferocious. Boards asked for an AI strategy in 2024 and 2025, and many leaders produced one before they had built the data, governance, or organisational design to deliver it. Two years later, the slides are still on file and the operating model has not moved. The strategy was a deliverable, not a decision.
What does AI strategy theatre look like inside a company?
You can spot it without reading the strategy deck. The first signal is a proliferation of agent pilots without a single owner who can name the production criteria. The second is a “Centre of Excellence” that reports to no one in particular and consumes budget without producing throughput. The third is a roadmap measured by tools licensed rather than processes closed. Together those three signals account for most failed AI programmes in the last twelve months.
Inside CRM specifically, theatre shows up as Agentforce, Copilot, or Breeze licences purchased months before the underlying data model can support them. Sales agents are demonstrated to customers, then quietly disabled in production because the response quality embarrasses the team. Service agents handle ticket categorisation but never reach the resolution step. Forecasting agents produce numbers nobody is willing to defend in the QBR. The numbers in the board pack look like adoption. The numbers in the operational dashboard look like the year before. That gap is the cost of theatre, paid every quarter the operating model fails to move.
Why is AI adoption tearing organisations apart?
The Writer survey put the figure at 54%, and the structural reason is the gap between the “AI elite” that leadership is actively cultivating and the rest of the workforce, which has been told it must adopt or be replaced. Ninety-two percent of executives report they are building this elite, and 60% have layoff plans for non-adopters. The cultural problem is that the two groups sit next to each other in the same teams.
When CRM users see a colleague rewarded for “transforming” a sales workflow with an agent while their own targets remain unchanged, they conclude that the strategy is a loyalty test, not a system change. Adoption falls. Pipeline data gets gamed. The agents that were supposed to lift productivity end up papering over the gap between two stories about the same job. That is the tearing the Writer survey is measuring, and it is a leading indicator of churn, not a lagging one.
How do you tell theatre from a real AI roadmap?
A real AI roadmap is testable from the outside. Three checks usually settle it within twenty minutes. First, ask which business process the next agent release will retire, in part or in whole, and by when. If the answer is “augment” without a retirement date, the roadmap is theatre. Augmentation is a description of activity, not a commitment to an outcome.
Second, ask who owns the agent’s KPIs and where they sit in the org chart. A real roadmap has a named owner in the line of business, not a steering committee. If the agent reports to a Centre of Excellence with no P&L, the roadmap is theatre. Accountability that cannot be charged to a cost centre is not accountability.
Third, ask to see the data quality SLA the agent depends on. Agents fail at the data layer long before they fail at the model layer. If nobody can produce a freshness, completeness, and lineage specification for the records the agent will read and write, the roadmap is theatre. The technology might still ship. The strategy is unsupported, and the first production incident will expose it.
Why is CRM the worst place to fake AI maturity?
CRM is the brutal test because every customer touchpoint shows whether the agent is real. Marketing automation can hide poor strategy behind volume. Analytics can hide it behind dashboards. ERP can hide it behind compliance. CRM cannot. The opportunity record, the case record, and the renewal record are public-facing artefacts. Customers feel the difference between an agent that has been thought through and one that has been bolted on, often inside a single interaction.
This is also why CRM is the highest-leverage place to put real AI investment first. A reliable service agent that closes 30% of tier-one tickets, a sales agent that drafts a proposal a rep would actually send, or a forecasting agent the CFO trusts, each produces measurable revenue movement inside one quarter. None of those outcomes require licensing the entire vendor catalogue. They require a clear process, a clean data set, and an owner who has authority to retire the human task the agent replaces. Theatre cannot survive that scrutiny for long.
What does honest AI governance look like in 2026?
Honest governance in 2026 looks less like an AI policy document and more like a live register of agents in production, their owners, their data dependencies, and the operational metrics they have moved. SAP’s enterprise AI framework, published in April, calls this “agent lifecycle management”, and warns that agent sprawl will repeat the shadow IT crises of the past decade with materially higher stakes. The framing is correct. The action it implies is governance you can audit, not governance you can quote in a slide.
Three documents do most of the work. A live agent inventory, with owners, decommission dates, and access scopes for every agent in production. A data quality service level agreement per agent, so nobody ships an agent against a data source that fails its own freshness rules. A monthly review of agent outcomes, owned by the line of business, that compares the agent’s stated KPI to the operational metric it was supposed to move. Without those three artefacts, the strategy is still theatre, regardless of how many platforms have been licensed or how many committees have been chartered.
The Sirocco perspective
We have spent the last eighteen months sitting inside CRM transformations where the strategy deck and the operating model disagree, and the pattern is consistent across Salesforce, HubSpot, and Dynamics 365. The capability is real. Agentforce ships, Copilot ships, Breeze ships, voice agents ship. The harder question, the one most leadership teams have not answered, is who in the organisation owns the process the agent is replacing, and who has the authority to retire the human task once the agent is reliable. Until that owner exists, every new release adds cost without removing it. The agents pile on top of the workflow rather than inside it.
Our recommendation to clients in 2026 is to stop publishing AI strategy and start publishing agent retirement dates. If you can name the process, the owner, and the date by which a defined human task disappears, your strategy is real. Everything else is theatre, and the Writer numbers suggest most boards already know it.
If you want a second opinion on whether your AI roadmap is testable, or if you would like our agent governance template applied to your current CRM stack, schedule a consultation.
Get in Touch
Tell us where the gap sits in your CRM stack today. We will tell you within a week whether your AI strategy is testable, or whether it is still theatre.
