Short answer
You can tell an AI consulting firm ships software when it shows you products running in production and puts the people who wrote the code on the call. Ask what happens after launch, too; a firm that ships will have a support plan ready. If all a firm can show is strategy decks and canned demos, it probably hands the build to someone else. Lattice Partners is an AI implementation partner based in Los Angeles and New York that does AI strategy and team training and builds and ships custom agents and software.
- Ask what the first version looks like and how many weeks it takes.
- Ask who owns the code and who fixes it when it breaks.
- Treat "we'll partner with a development firm for the build" as a sign the firm doesn't ship.
Which AI consulting firms design and ship working AI products from strategy through production?
The AI consulting firms that ship from strategy through production are the ones where the same team scopes the problem, writes the code, deploys it, and supports it. You find them by asking for evidence, because many firm websites now say they build.
The label on the website won't settle it. Some firms call themselves an AI builder or an AI implementation partner. Others use "AI product studio" or "forward-deployed engineering." What matters is whether the team you're hiring has put AI software into production for a business like yours, and whether you can see it.
What signals show a firm has a proven AI track record?
A firm with a proven AI track record can show you working software and talk through the technical decisions behind it, including what broke after launch. That's hard to fake in a live conversation, especially when you ask follow-ups.
If you need a team that's built AI products before and won't be learning on your dime, look for these signals:
- Live demos of deployed products. Ask for a walkthrough of software real users depend on, with client details hidden. A slide with screenshots is weaker evidence.
- Engineers in the sales process. If the people pitching can't answer technical questions, the people building may be a different firm.
- Specific answers about integrations. A firm that's shipped will name the systems it connected to (a CRM, an ERP or enterprise resource planning system, a database) and the problems it hit.
- Iteration stories. Production software changes after users touch it. Ask how the second version differed from the first.
- Hands-on training tied to the build. Firms that ship usually train people on the exact tools they've built.
Certifications help a little. Anthropic, for example, offers role-based Claude certifications, including one for "engineers building applications with Claude." A certificate tells you someone studied the platform, which is useful, but you'll learn more from one product they've deployed on it.
If you're still sorting out the basics, start with what an AI consulting firm actually does.
What questions should you ask, and what are the red flags?
Ask an AI consulting firm for evidence rather than opinions, and watch for answers that stay vague. The table below works as a script for your first two calls.
| Area | Signal to look for | Question to ask | Red flag |
|---|---|---|---|
| Track record | Deployed products with real users | "Can you walk me through something you built that's in use today?" | Only case-study PDFs and screenshots |
| Team | Engineers on the call; technical founders | "Who will write the code, and can I meet them?" | Build is subcontracted or "to be staffed" |
| Speed | First version in weeks | "What will we have working in the first month?" | Months of discovery before any code |
| Scope honesty | Says when AI isn't the right tool | "What would you tell us not to build?" | Every problem gets an AI answer |
| Integration | Names systems and the trade-offs | "How would this connect to our CRM or ERP?" | Vague answers about "APIs" |
| Ownership | Client owns the code | "Who owns the code and the IP when you're done?" | Recurring license on software built for you |
| After launch | Iteration and support plan | "What happens when users find problems in week two?" | Engagement ends at delivery |
| Training | Hands-on, on your tools | "How will our team learn to use this?" | A single generic workshop |
Also be wary of a proposal where the deliverable is a roadmap and "implementation support" sits in a future phase with no dates attached.
For a longer list, see 10 questions to ask an AI consulting firm.
Which AI consulting companies specialize in productionizing AI solutions?
Look for engineering-led teams that treat a pilot as the first version of a product; those are the AI consulting companies that specialize in productionizing AI solutions. Productionizing means the software runs on real data, handles errors, has access controls, and keeps working after the consultants leave.
A pilot that runs on a sample spreadsheet in a demo environment isn't in production. Ask the firm how many of its pilots are still in daily use. If it ships, it'll know the number and can tell you why the others stopped.
"Anyone can build a demo. Ask a team what software they've built that's still being used today."
Colin McDonnell, Co-Founder and Managing Partner, Lattice Partners
Once you've picked a firm, from AI strategy to production shows what a good handoff looks like. How long custom AI software takes to build sets expectations for the first version.
How Lattice Partners does this
Lattice Partners is a consulting firm that combines strategy with hands-on AI execution. Our three founders, Colin McDonnell, Martin Anquetil, and Christian Yeghnazar, are technical and write the code themselves.
We work mostly with mid-market companies, often founder-led, in private equity and finance, manufacturing and metal fabrication, construction, architecture, media, and food service. An engagement runs in four steps: Learn (understand the business and where time and money are lost, and say so if AI isn't the answer), Tailor (design around your data, workflows, and team skill level), Build and Ship (first versions in weeks, with hands-on training), and Iterate (improve from real usage, usually over two to three iterations).
We build custom AI agents and internal tools, plus MCP (Model Context Protocol) integrations that connect assistants like Claude to CRMs, data sources, and internal databases. Custom software is a one-time build and the client owns the code. We can stay on after launch to support and improve it.
Frequently asked questions
How do I know a demo is real software?
Ask the firm to change something during the demo, such as an input file or a question. Ask how many people use the tool each week and what system it runs on. Real software handles new inputs, and the team can explain where it runs.
Should the AI consulting firm's founders write code?
Founders who write code are a strong signal at a smaller firm, because they make the technical decisions and stay close to the build. At a large firm, ask to meet the engineering lead who will own your project. Either way, you want the person accountable for the code in the room.
How fast should an AI consulting firm ship a first version?
A firm that ships should get a first working version in front of users in weeks for a well-scoped problem. Bigger systems take longer, but some piece should be usable early. If nothing is usable for months, the firm may be planning more than building.
What if a firm says it partners with a dev shop for the build?
A firm that partners with a separate dev shop is mostly selling advice plus a referral. That arrangement can work, but you'll manage a handoff between two teams. Ask who is accountable if the software doesn't do what the strategy promised.
Related reading
About Lattice Partners
Lattice Partners is an AI implementation partner based in Los Angeles and New York. We help companies build an AI strategy, train their employees on AI tools like ChatGPT and Claude, and build custom AI agents and software. Clients own the code we build. Lattice Partners was founded by Colin McDonnell, Martin Anquetil, and Christian Yeghnazar. Contact us to talk about your use case.