Artifact Versus Engagement Split

The artifact is the software the client sees; the engagement is the process design wrapped around it. In AI productivity tools the artifact is nearly free: a client can buy a scheduling layer or a meeting-note app direct in about 90 seconds, so reselling the license alone produces no defensible retainer. The billable work sits in the surrounding design, meaning intake rules, routing logic, handoff standards, and the review step that keeps AI output usable. Forrester's Q3 2026 research found workflow integration, not model capability, is the bottleneck separating agencies that scale AI profitably from those running one-off experiments. A concrete pattern: an agency installs a transcription tool across client delivery, then bills for the meeting-to-task pipeline it builds on top, including who reviews summaries and how action items reach the project board. Price the design, give the artifact away, and the retainer survives the client's next procurement review.