Rebuilding Business Capability with AI

AI drafts a client proposal in ten minutes. At the review meeting, nobody can say which assumptions were checked or which commitments the team can make. The document took less time to write, but the client is still waiting. To judge whether AI improved operations, follow the proposal through approval, delivery, and use.
Start with the people who use the result
Consider a firm preparing a bespoke quote. The client needs a quote they can make a decision on. The delivery team needs an accurate scope, and finance needs to understand the costs and risks. Write down what each party will use to decide before assigning work to AI.
The same output may serve several people, each with different criteria. A client needs a clear price. Finance needs to trace the calculation. Someone must have the authority to resolve conflicts between those requirements.
Break the business outcome into work that can be handed off
Andrew David W.Y.’s exploratory V2.0 framework outlines a four-direction SOP. For a quote, trace the outcome from business performance through the workflow and the functions involved down to individual actions. Then trace it back up. Which risk does checking a particular input reduce, and whose next decision does it support?
Business leaders decide which client problems to address. Process owners set responsibilities and authorizations. Practitioners deliver the work and record exceptions. AI can extract requirements, compare line items, and organize evidence. An authorized person remains accountable for commitments on price, scope, and timing.
Ask two questions at every handoff
First, does the previous step meet its acceptance criteria? Return a quote for rework when source data or cost evidence is missing. Second, can the next step start? An acceptable quote may still lack approval, budget, or confirmation from the client. Record the blocker and assign it to the person who can resolve it.
One handoff record can capture the deliverable ID, reviewer, version of the standard, decision, and outstanding dependency. Those fields help teams distinguish a quality problem from a coordination problem.
Make the improvement usable next time
Check four things separately: whether the work happened, whether the output passed review, whether the next team used it, and whether business results changed within the agreed period. Compare first-pass acceptance, rework, waiting time, and cost to see whether a faster step simply moved work downstream.
When a change works, put the same acceptance criteria in role descriptions, human work instructions, performance review rules, and AI skills. Record the scope, owner, and version. Ask a colleague outside the original pilot to deliver similar work independently using the new standards.
V2.0 offers a way to connect stakeholder value, accountability, and evidence. The quote is an illustrative example. Whether the approach reliably improves operations requires real data and repeated trials.
Research reference: Andrew David W.Y., Rebuilding Enterprise Operating Capability with AI (exploratory V2.0).

