Separate deliberation from delivery in AI initiatives
Committees deliberate well and deliver poorly; once delivery is the bottleneck, separate the functions and move delivery to a small dedicated cell.
The standard structural response to “we should do something about AI” is an AI committee — a cross-functional group that meets periodically, reviews progress, identifies opportunities and acts as the firm’s central point of co-ordination. The structure looks sensible. It is also reliably the wrong shape once delivery is the bottleneck.
A committee does two different things. It builds literacy across the firm by exposing members to AI use cases and tools. And it is asked to deliver concrete things — internal applications, knowledge bases, platform decisions. It is reliably good at the first and reliably poor at the second. Deliberation has a cadence; delivery has a deadline; the cadence wins.
The heuristic is to separate the functions when the second one stalls. Leave the committee in place with its literacy and communications role explicitly named. Move delivery work to a small cell — three or four people — staffed for build rather than discussion. The cell reports back to the committee on what it has built; the committee continues to surface use cases and educate the firm. Both functions get a body that fits them.
Three conditions usually trigger the pivot.
The first is Unvoiced staff resistance is the primary failure mode of AI initiatives showing up in committee dynamics — energy displacing onto adjacent problems instead of engaging with the work. The second is sponsor frustration accumulating: a senior stakeholder whose KPIs depend on visible progress, watching a deliberative body produce conversation rather than tools. The third is a specific delivery item — a knowledge base, a billing tool, a platform decision — that has been on the agenda for months and has not advanced.
The case study A mid-tier firm’s pivot from AI committee to delivery cell works through one engagement where all three conditions were present. The pivot was unilateral on the sponsor’s part, after a frank exchange, and unblocked work that the committee structure had been unable to produce.
The committee structure is not wrong. It is asked to do too much.