AI strategy consulting engagement deliverables, plus the data asset review most skip

What should an AI strategy engagement deliver?

A typical AI strategy consulting engagement delivers five things: a current-state assessment, a ranked portfolio of use cases, a data and technology readiness review, a governance and risk outline, and a phased roadmap with owners and budgets. Most engagements stop there. The gap is that every one of those deliverables asks what the client can do with AI, and none asks what AI developers might pay the client for.

That reverse question deserves its own short workstream, because the answer can change how the roadmap is funded.

Why consultants are well placed to ask the reverse question

You already hold what the review needs: executive access, a mandate to inventory systems, and interviews with the people who run them. In the first two weeks of current-state work, a consultant sees the CRM, ticketing, ERP, engineering and document systems and hears which of them go back a decade.

The question is also timely. Researchers at Epoch AI estimate the stock of public human-written text at roughly 300 trillion tokens and project that language models could fully use it between 2026 and 2032 if current trends continue. It is a forecast with wide uncertainty, but it helps explain why AI developers look for licensed, non-public records of real work: multi-step workflows, decisions and outcomes that rarely appear on the public web. For who those developers are and what they buy, see what an AI data buyer is.

Which clients in your book fit

Industry matters less than how the work is recorded. Look for these signals in engagements you already have:

B2B software, IT services and managed service providers, professional services, engineering, logistics and distribution clients tend to score well. The who qualifies page sets out the full baseline.

The RACE check: a 30-minute screen inside the engagement

Run it during current-state work. A clear no on any line parks the topic.

• Records: do several core systems hold five or more years of history, and were retired systems archived rather than deleted?
• Authority: is there an owner, CEO, CFO or authorized representative who could sponsor a licensing decision?
• Clean rights: did the client create the records, and do its client contracts, privacy policy and employee notices leave room to license them?
• Exportability: can someone at the client, or its IT provider, actually export each system?

Scoping the data asset review as a separate workstream

Keep it short, separate from the main deliverables, and limited to metadata. One consultant can usually cover it in two to three weeks of part-time effort alongside current-state work:

1. System census. List each system with its business owner, earliest year of records, approximate volume and whether exports are possible. Record names and counts, never contents.
2. Rights read. With the client's counsel, flag client contracts that restrict use of communications or deliverables, privacy promises, recording notices and any earlier data license.
3. Outcome scan. Mark which systems tie work to results, such as resolution codes, win and loss reasons, or approval decisions.
4. Finance note. Tell the CFO that the structure of a data license can affect when revenue is recognized. Deloitte's ASC 606 roadmap explains the right-to-use versus right-to-access distinction for licenses of intellectual property; the client's auditors decide how any particular license is treated. This is general information, not legal, tax or financial advice.
5. Decision memo. Two pages: what the review found, what is out of bounds, and one decision for the owner on whether to request a SourceX introduction.

The memo is the deliverable. Do not value the records, quote a price or suggest buyers. Price is agreed between the company and SourceX, and nothing is binding until the company signs.

When to raise it in the engagement calendar

Some consultants seed the topic months earlier by hosting a client roundtable on AI and company data, then propose the workstream to the owners who lean in.

How the introduction works once the owner says yes

1. You register as a partner, then send the owner your referral link or submit the company through the referral form with the owner's permission.
2. SourceX checks size, history, breadth of records and rights with the owner or their delegate.
3. The client completes its own data inventory with SourceX; your census can help the client's team, but you hand nothing confidential to SourceX.
4. SourceX and the client agree one all-in price and the license terms, typically an exclusive AI-training license for an agreed term.
5. AI labs and data buyers review the opportunity, and the client signs only if the terms work.
6. Records are prepared and delivered under de-identification and redaction rules agreed before any work begins, and the client is paid.

Keep your two roles apart. Inside the engagement you may see confidential records; as a referral partner you pass on only basic fit information.

What to say at the readout

How rewards work for consultants

Partners earn 25% of the eligible platform fees SourceX actually collects from the referred company's licensing deals, capped at $100,000 per referred company. Payment follows only after the buyer pays and SourceX receives its fee. The reward comes out of SourceX's fee rather than the client's proceeds, and no reward is guaranteed.

Because the client pays you for advice, tell the client about the referral relationship in writing before you introduce them, check your engagement letter and firm policy for conflict terms, and keep the recommendation independent of the reward. The management consultants partner page goes further into how consultants handle this.

When to leave it out of the engagement

• Peak full-time headcount never reached 50, or the operating history is short.
• Most records belong to the client's own customers, as at agencies and outsourcers, and those customers have not agreed.
• The material is mainly consumer personal information or protected health information.
• Archives were deleted when systems were replaced, or nobody can export them.
• The same records are already licensed for AI training.
• The owner will not consider an exclusive license for an agreed term.

Next step

Add the RACE check to your next current-state template, and use the network opportunity finder to look back across past engagements for clients worth a second conversation. When an owner says yes at readout, register as a partner and make the introduction.