Where should AI land first?
Wherever the workflow says, not wherever the demo lands. A 1–2 week assessment that maps the difference.
For CEOs and GMs deciding where AI should land first.
Scope fixed before kickoff. Pricing set at the discovery.
Most AI rollouts that fail were misrouted at the start. The team that picked the platform was not the team that did the work. The pilot that launched was not the workflow most worth changing. The Discovery exists to fix the order: map your team's actual work before the platform decision is made.
Licensing AI and using it are not the same thing. In an independent survey of employees with Microsoft 365 Copilot access, just 35.8% actively use it.1 The Discovery surfaces who is actually using it, which workflows fit them, and which ones need different tools or different change support.
1 Microsoft 365 Copilot workplace conversion rate: the share of employees with workplace access who actively use it. Source: Recon Analytics, AI Choice 2026: Why Licenses Don't Equal Adoption (February 2026), an independent survey of 150,000+ U.S. respondents.
In a three-week randomized trial at Microsoft, developers given GitHub Copilot rated it more useful, while telemetry showed no statistically significant change in their output.6 The Discovery measures the workflow that ships work, not the vendor's productivity claim.
Two scopes
Founder's Edition
1 week, single function. Built for SMEs and founders who want a clear answer on one workflow.
1 week · single function
Standard
2 weeks, multi-function. Built for organizations evaluating where AI should land across two or more functions.
2 weeks · multi-function
What you get
A workflow map. Each task sorted into three buckets (AI absorbs, AI supports, AI stays out) with a change-effort estimate for each.
Frameworks applied
AI risk standards — NIST AI RMF and ISO/IEC 42001 frame the risk read on every task, flagging where regulated data, customer decisions, or compliance-critical processes change the answer.
Workflow analysis methodology — The AI-Workflow-Fit Diagnostic sorts each task into absorb, support, or stay-out, with Lean Six Sigma DMAIC and Value Stream Mapping principles underneath the mapping for ops-heavy workflows.
Cynefin Framework — The sense-making layer beneath the three categories. Clear and Complicated tasks are where AI absorbs, Complex is where it supports, and Chaotic or high-stakes Complex is where it stays out.
Jobs to Be Done — Sharpens the workflow read with one question: what job is the worker actually hiring AI to do? It resists fashion-driven automation and surfaces the jobs people care about but rarely articulate.
6Butler, J., Suh, J., Haniyur, S., & Hadley, C. Dear Diary: A randomized controlled trial of Generative AI coding tools in the workplace. arXiv:2410.18334, October 24, 2024 (preprint). A three-week trial at one software company: 269 engineers enrolled, 106 analysed. Developers given GitHub Copilot rated the tools more useful and enjoyable; telemetry (code changes, pull requests, pull-request hours, development time) showed no statistically significant difference, and the authors note the trial may have been too short.