AI lab, consulting firm or software vendor: which should run an enterprise AI project?
A software vendor sells a tool, a consulting firm sells a plan, and an AI lab like HomoDeus takes a hard problem and builds the agent that solves it on the company's own records. The right choice depends on whether the problem is common or specific to the company.
Why does the choice matter?
According to RAND's 2024 report, by some estimates more than 80% of AI projects fail, twice the rate of IT projects without AI. A tool leaves the result with your team, and a plan leaves it with whoever executes it. A lab builds the agent on your records, so the test is whether it works there.
Side by side
| Software vendor | Consulting firm | AI lab (HomoDeus) | |
|---|---|---|---|
| What you get | A product your team adapts to | A diagnosis and a roadmap | An agent working on your records |
| Whose knowledge it runs on | The vendor's general model | What the team gathered in interviews | Your people's judgment, captured |
| Best fit | A common problem many companies share | A question of strategy or change | A hard problem specific to your operation |
| Who signs the outcome | Depends on the product | Your team, after the project | A person, on every risky outcome |
Questions
When is a software tool the better choice?
When the problem is common and a tool already fits how the team works. At the US brokerage behind Scribe the off-the-shelf tools could not do the work, so a team of agents was built for it.
What does an AI lab need from the company?
Access to the records and time with the people who hold the judgment. Keystone was rebuilt alongside the desk it serves.
Do the agents replace the team?
No. In the HomoDeus cases the agent takes the reading and the routine, and a person signs where the risk sits: the partner, the manager, the compliance officer, the editor.