Do you need an AI consultant, or can you handle AI internally?

You can handle AI internally when you already have two things: a clear picture of how the work really runs, and someone who has shipped AI into daily use before. Many mid-sized companies have neither, and those two things, not the technology, are what outside help should supply. If what you need is a decision (which business to exit, how to restructure a team), a consultant is the right buy. If what you need is software people use every day, a plan is only the start, and you want whoever helps you to build it with your team and hand it over.

How do you decide between a consultant, a partner and your own team?

"Consultant or in-house" is usually asked as a question about cost. It is more useful as a question about gaps. An AI project in a company of a few hundred people needs four things:

1. A picture of how the work really runs. Not the process document; the route the work actually takes, with its workarounds.
2. A way to choose what to build first, ranked by what each problem costs today.
3. Someone who has shipped AI into daily use before, and knows where it breaks.
4. Someone who will run it afterwards.

Whoever supplies the ones you lack is the help you need. Most companies have the fourth (their own team). The first and third are the two that decide whether you can go it alone.

Do AI projects fail more often when you build them in-house?

A widely cited 2025 study, MIT NANDA's The GenAI Divide, found that about 5 percent of AI pilots achieved rapid revenue acceleration while the rest stalled. It also found that "purchasing AI tools from specialized vendors and building partnerships succeed about 67% of the time, while internal builds succeed only one-third as often" (Fortune, 2025). Note what "partnership" means there: buying from specialist vendors or building with them, not hiring advisers. The sample is also modest: 150 interviews, a survey of 350 employees and 300 public deployments.

The skills gap is the reason most often given. Among EU enterprises that considered AI and decided against it, 71 percent cited a lack of relevant expertise (Eurostat, 2025).

Neither route is safe by default, though. BCG's 2024 survey of 1,000 executives found 74 percent of companies had yet to show tangible value from AI (BCG, 2024). Gartner expects more than 40 percent of agentic AI projects to be cancelled by the end of 2027; its analyst described most current projects as "early-stage experiments or proof of concepts that are mostly driven by hype and are often misapplied" (Gartner, June 2025, as reported by RCR Wireless).

When to keep it in-house

• You have an engineer who has already put an AI tool into daily use, not only built a demo.
• The process you want to change is well understood and runs mostly inside one system.
• You can give that person time: an AI build done in the margins of a day job rarely ships.

In that case, buy the tools you need and spend on your own people. A partner would mostly be renting you what you already have.

When to bring in a consultant

• The deliverable is a decision, not software.
• You need an outside view the board will accept.
• You need specialist judgement for a short time, such as a data-protection or regulatory review.

Know what you are buying: a strategy engagement usually ends at the recommendation, so if the plan does not fit how the work actually runs, you find out after they have left. Many consultants also build; if yours does, judge them as a building partner, with the questions below.

When to bring in a building partner

• Nobody inside can describe how the work really runs end to end, because it lives with the people doing it.
• You want something in daily use this year, and nobody on the team has shipped one before.
• You want your own team to own it afterwards.

Look for a partner who starts by finding out how the work happens, builds inside the systems your people already use, works alongside your team rather than behind a wall, and hands everything over, code included. Know what you are buying here too: your team has to give real time to the build, and if nobody on your side can, the handover will not stick.

What should you ask an AI consultant or partner before hiring them?

1. What will you find out about how our work runs before you build anything, and how?
2. Who decides what gets built first, and on what numbers?
3. Who builds it, and who from our side sits in the build?
4. What do we own when you leave: code, data, documentation?
5. What happens in the month after it ships?

If the answers are vague on the first question, the rest will be built on the documented process, and the documented process is not the one running.

Our view

In our experience the missing piece is almost never the technology. It is the picture of how the work really runs, which lives with the people doing it and rarely makes it into a system or a process document. That is why we built Transcript, which talks privately with the people who do the work and builds one picture of how the work happens. Our engineers then build with the company's own team. If you already have that picture and someone who has shipped before, keep it in-house.

Questions people ask next

Do I need an AI consultant if I have an IT team?

Not necessarily. An IT team knows your systems; what it often lacks is time and a picture of how the work runs outside those systems. If your IT team has both, it can lead. If not, bring in help for the discovery and the first build, and have your IT team build alongside so they can run it afterwards.

When should I not hire an AI consultant?

When you already know exactly what to build and have someone who can build it, or when the problem is small enough to solve with one good tool and one changed process.

Is it cheaper to build AI in-house?

Payroll looks cheaper than a partner until you count the time before anything is in daily use. MIT's finding that internal builds succeeded about a third as often as bought tools and partnerships (roughly 22 percent against 67) is the number to weigh against the outside rate.

What should an AI partner hand over at the end?

Everything: the code, the systems, the data and the documentation, running in your environment, so your team can run it without them.

References

• Fortune, MIT report: 95% of generative AI pilots at companies are failing, reporting on MIT NANDA, The GenAI Divide: State of AI in Business 2025 (2025), https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/
• Eurostat, Use of artificial intelligence in enterprises (2025), https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises
• BCG, AI Adoption in 2024: 74% of Companies Struggle to Achieve and Scale Value (2024), https://www.bcg.com/press/24october2024-ai-adoption-in-2024-74-of-companies-struggle-to-achieve-and-scale-value
• RCR Wireless, Gartner: More than 40% of agentic AI projects will fail by 2027 (27 June 2025), reporting Gartner's press release of 25 June 2025, https://www.rcrwireless.com/20250627/business/agentic-ai-gartner