What questions should I ask my mobile vendor to assess their AI capability?

Ask four questions. First: name an AI feature you have shipped in a production mobile app - not a prototype, a live feature with real users. Second: show me an example of an AI workflow output from a recent project - a code review comment, a release note, a regression result. Third: describe how you handled the privacy review for an AI feature in a regulated environment. Fourth: who on your current team has AI integration experience, and what have they built? A vendor who answers all four with specifics and evidence has demonstrated genuine capability. A vendor who generalizes or cannot produce artifacts has not.

How long does a vendor AI capability assessment take?

The conversation itself takes 60 to 90 minutes. Reviewing any artifacts the vendor provides takes another two to four hours. The proof-of-concept phase - if you run one - takes six weeks. The full assessment from first conversation to go or no-go decision is typically two to three weeks if you run it promptly. Delaying the assessment because the vendor relationship is comfortable costs time you cannot recover if the board deadline is real.

What should I do if my vendor passes the conversation but I still have doubts?

Run the proof-of-concept. A six-week phase with a specific deliverable - a working prototype of the first AI feature in a test environment, meeting a defined accuracy and latency threshold - is the most reliable test available. Conversation performance and delivery performance are not the same. The proof-of-concept closes the gap between what the vendor says they can do and what they actually deliver under real project conditions.

Can my vendor add AI capability by hiring or partnering?

Yes, but verify before you accept it as a plan. A vendor who says they will hire an AI engineer for your project is making a promise about future capability, not current capability. Ask who specifically will be hired, what their background is, and when they start. A vendor who says they will partner with an AI specialist should be able to name the partner and describe the working model. Vague plans to add AI capability through future hiring or unnamed partnerships are a risk factor, not a mitigation.

Writing

How to Know If Your Current Vendor Can Deliver Your Board AI Mandate

Before you hand an AI mandate to your existing mobile vendor, run four checks. The answers tell you whether to proceed, add a specialist, or find a new team entirely.

Ali Hafizji · CEO & Co-founder, Wednesday Solutions

7 min read·Published Mar 29, 2026·Updated Apr 26, 2026

4xfaster with AI

2xfewer crashes

10xmore work, same cost

4.8on Clutch

Trusted by teams at

In this article

The four checks
Check one: shipped AI reference
Check two: workflow evidence
Check three: compliance track record
Check four: team composition
What the results tell you
Frequently asked questions

The board gave you an AI mandate. Your first instinct is to hand it to the team already working on your app. They know the codebase. The relationship is established. The onboarding cost is zero.

Before you do, run four checks. Each one takes under an hour. Together they tell you whether your current vendor can deliver the mandate, needs support to deliver it, or is not the right team for this project.

The cost of skipping the checks is a six-month project that stalls at month three when the capability gap becomes visible. Four hours of assessment now versus three months of delay later is the trade-off.

Key findings

Vendor AI capability is not self-reported accurately. Vendors who cannot deliver AI features consistently describe themselves as AI-capable because they use AI tools in their workflow. The distinction between using AI tools and having delivered AI features in production apps is the one that matters - and it is testable in a single conversation.

The checks are structured to produce a yes, a conditional yes, or a no. A conditional yes - the vendor has partial capability and needs support - is a useful outcome. It tells you exactly what needs to be added rather than leaving the question open.

Running the assessment before committing the budget is the normal practice in well-run technology organizations. If the vendor relationship makes you reluctant to run it, that reluctance is a signal worth examining.

The four checks

The four checks test four distinct capability areas. A vendor can pass three and fail one. The pattern of passes and fails tells you more than the overall score.

Check one tests delivery track record. Check two tests workflow maturity. Check three tests regulated environment experience. Check four tests team composition. A vendor who passes all four is a strong candidate for the AI work. A vendor who passes one or two has a gap that needs to be filled before the project starts.

Check one: shipped AI reference

Ask the vendor to name a production AI feature they have delivered in a mobile app. Not a prototype, not an internal tool - a live feature with real users, currently running in an app on the App Store or Google Play.

The reference should be specific enough that you can download the app and use the feature. If the vendor names a client who they cannot name publicly, ask for the feature description and the metrics it produces. A vendor with genuine AI delivery experience can describe the feature, the model it uses, the latency in production, and at least one metric that shows it is performing.

A vendor who cannot name a shipped AI reference has not delivered AI in production. Their next project for you would be their first.

Check two: workflow evidence

Ask the vendor to show you an AI workflow output from a recent project. Not a description of their workflow - an actual output.

Three artifacts work for this test: an AI-generated code review comment from a recent release, a sample of AI-generated release notes from a shipped version, or a screenshot regression comparison showing AI-detected visual differences between builds. Any of the three demonstrates that AI is integrated into the day-to-day delivery process, not just claimed as a capability.

A vendor who can produce one of these artifacts in the conversation is using AI operationally. A vendor who describes their AI tools without being able to show an output is not.

Check three: compliance track record

If your app operates in a regulated industry - financial services, healthcare, edtech with minors' data - ask specifically about their experience navigating AI-related compliance requirements.

The questions: Have you shipped an AI feature in a HIPAA, SOC 2, or FINRA-regulated environment? How did you handle the privacy review for the AI data flows? What did the App Store review process look like for the AI features? Have you worked with a CISO on an AI feature approval?

A vendor with genuine regulated industry AI experience can answer each of these specifically. A vendor without it will answer in general terms about compliance practices. General terms are not wrong - they are insufficient for a regulated environment project.

Case study — Clinical digital health platform

0patient logs lost offline — seizures logged anywhere, synced automatically

“They really cared and felt like an extension of our team. The quality of the work was top notch, and they were receptive to shifting priorities.”

Founder, Digital health platformRead the case study →

If you are assessing whether your current vendor can deliver an AI mandate and want a structured way to run the conversation, a 30-minute call covers the framework and what to look for in the answers.

Book my call →

Check four: team composition

Ask who specifically would work on the AI features. Not the team in general - the individual engineers who would own the AI integration, model selection, and AI QA.

Ask what each person has built in the AI space. Ask whether they have worked on on-device AI or only cloud-connected features. Ask whether there is an AI specialist on the team or whether AI work would be distributed across generalist engineers.

A vendor with genuine AI capability can name specific team members with specific AI experience. A vendor who answers this question in the abstract - "our team has strong AI expertise" - has not actually assigned anyone with that expertise to your project.

What the results tell you

Four passes: proceed with the vendor for the AI work. The track record, workflow, compliance experience, and team composition are all in place.

Three passes, one fail: proceed with a structured proof-of-concept before committing the full budget. The gap is specific and testable. If the proof-of-concept meets its success criteria, the fail was a minor capability gap. If it does not, you have six weeks of investment rather than six months.

Two or fewer passes: the vendor is not the right team for the AI mandate. The options are to bring in a specialist vendor to work alongside them, or to evaluate a new vendor for the full engagement. Either path is better than proceeding on the assumption that the gaps will close during the project.

Wednesday steps in when existing mobile vendors reach their AI capability limits. A 30-minute call covers what that transition looks like and how to structure the engagement.

Book my call →

Frequently asked questions

The writing archive has vendor comparison guides, cost benchmarks, and decision frameworks for every stage of the enterprise mobile buying process.