Which Evidence to Request from an AI Consultancy: The List of 12 Documents Before Deciding
By Pedro Sonda, dooop
Before hiring an AI consultancy, request evidence in four groups: capability (detailed cases and references), method (methodology and examples of deliverables), team (names, roles, dedication, and resumes), and security (data handling, confidentiality, and the consultancy’s own use of AI). Requesting the same set from all candidates allows you to compare facts, not just impressions from meetings.
Updated October 2026.
Transparency note: dooop is an AI consultancy. This list applies to any provider, including dooop itself.
Why Request Documents, Not Just Conversations
Commercial meetings are good for assessing understanding and rapport but poor for comparing providers. Each tells their own story with different emphases. A standardized request for evidence creates a common basis and quickly reveals who has concrete material and who only has rhetoric.
Gartner identified inadequate risk controls and unclear business value among the reasons that at least 30% of generative AI projects were projected to be abandoned after proof of concept (Gartner, 2024). Both points can be verified before hiring if you request the right evidence.
dooop Framework: The 12 Pieces of Evidence in 4 Groups
Group 1: Capability
| # | Evidence | What It Shows | What to Observe |
|---|---|---|---|
| 1 | Two to three detailed cases | Experience with similar problems | Problem, approach, result, and limitations |
| 2 | Client references | External confirmation | At least two, from comparable projects |
| 3 | Published content | How the consultancy thinks | Articles, research, or talks about your type of problem |
Group 2: Method
| # | Evidence | What It Shows | What to Observe |
|---|---|---|---|
| 4 | Described methodology | How the work happens | Steps, decisions, and your team’s participation |
| 5 | Example of a real, anonymized deliverable | Quality of what you will receive | Whether it ends in decisions or just analyses |
| 6 | Proposed success indicators | How results will be measured | Business indicators, not activity metrics |
Group 3: Team
| # | Evidence | What It Shows | What to Observe |
|---|---|---|---|
| 7 | Names, roles, and dedication | Who will actually work | Percentage of time each person dedicates |
| 8 | Summarized resumes | Individual experience | Combination of business, change management, and AI expertise |
| 9 | Substitution rule | Continuity | Who approves team changes |
Group 4: Security and Compliance
| # | Evidence | What It Shows | What to Observe |
|---|---|---|---|
| 10 | Data handling policy | How your data will be used | Roles of controller and operator as defined by LGPD |
| 11 | Confidentiality agreement | Protection of strategic information | Scope, duration, and destination of information at contract end |
| 12 | Statement on the use of AI in deliverables | Whether and how the provider uses AI in their own work | Tools, data sent, human review |
Why Group 4 Deserves Special Attention
An AI consultancy will have access to processes, data, and often the company’s strategy. The LGPD defines responsibilities for those processing personal data, including the roles of controller and operator (Law 13.709/2018). The contract and the provider’s data policy must clearly state who is responsible for what.
Evidence 12 is recent and rarely requested. In 2025, Deloitte agreed to refund part of a contract with the Australian government after a delivered report contained citations and references with errors attributed to generative AI use; the revised version disclosed the use of the tool (The Register, 2025). This case shows that asking how the provider uses AI and reviews their deliverables is a matter of quality, not distrust.
How to Request the Evidence
- Send the same list, in writing, to all candidates on the shortlist.
- Accept that some cases may be anonymized or presented only in meetings due to confidentiality.
- Give the same deadline to all.
- Record what each delivers in a single table; see How to Compare AI Consultancy Proposals.
Interpretation: the absence of evidence is information. A provider unable to show any deliverable, even anonymized, or who cannot explain how they handle data, is signaling something important.
What to Do Now
- If you are creating the shortlist: send this list along with the briefing.
- If you have already received proposals: check which of the 12 pieces of evidence each includes.
- If you are about to sign: confirm that evidences 9 to 12 are reflected in the contract; see AI Consultancy: 20 Questions Before Signing a Contract.
To receive dooop’s evidence, contact is at dooop.com.br/contato.
Frequently Asked Questions
Is it reasonable to request all this for a small project? For small projects, items 1, 5, 7, 10, and 12 usually suffice. The rest can be verified in conversation.
What if the provider cannot show cases due to confidentiality? Request an anonymized presentation in a meeting and speak with references. Confidentiality explains the absence of names, not the absence of details.
Also Read
About dooop
dooop is a consultancy based in Caxias do Sul (RS) that prepares leaders and organizations to transform artificial intelligence into business capability through strategy, governance, and training.