Once you know which workflow to improve, there is a second decision that gets less attention: who does the work. You can buy a finished product, build something custom or bring in help to configure and manage it. Each path has a different price, timeline and way of going wrong.
What the research suggests
The MIT NANDA report summarized by Fortune found that tools bought from specialized vendors, and partnerships, succeeded about 67% of the time, while internal builds succeeded roughly one-third as often.[1] The summary does not explain why, and "success" was defined in terms of rapid revenue acceleration, so a modest but real gain could still be counted as a miss. Even so, it matches a common pattern: custom building is slower, harder to maintain and easier to underestimate than buying.
The three paths
| Path | Best when | Watch for |
|---|---|---|
| Buy a product | A mature tool exists for your exact workflow | Features you pay for and never use; weak integrations with your systems; lock-in |
| Build custom | Your process or data is unusual and a product cannot handle it | Maintenance falling on one person; scope creep; no one left who understands it |
| Hire help (consultant, fractional lead, implementer) | You lack the time or in-house skills to choose, set up and manage the tool | Unclear deliverables; dependence on the consultant; advice with no one accountable for results |
These can be combined. A common pattern is to buy the core tool and hire help to configure it, connect it to your systems and set up the approval steps.
Questions that decide it
- Does a product already do this? Search before you build. Our tools directory lists more than 700 options by category.
- Is the workflow standard in your industry? Standard workflows (invoice capture, call answering, meeting notes) are well served by products. Unusual ones are not.
- Who will maintain it in two years? If the answer is one employee or one contractor, you carry real key-person risk.
- How much of your data does it need? The more it touches, the more security review it deserves.
- What does it cost to leave? Can you export your data and configuration if you switch?
Total cost of ownership
Compare options on a three-year basis, not on the first invoice. Include license fees, setup, integrations, training, review time, ongoing changes and the cost of switching later. A cheap tool that needs a person to babysit it daily is not cheap.
Watch for agent washing
When you buy, you are buying a claim. Gartner coined the term "agent washing" for products rebranded as agentic AI without real capability, and estimates that only about 130 of thousands of vendors claiming it are genuine.[2] Ask vendors for a live test on your data, a list of exactly what the system is allowed to do without approval, and references from a company your size.
Vendor due diligence checklist
- Where is our data stored, and who can see it?
- Is our data used to train your models? Can we opt out in writing?
- What admin controls, logs and user permissions exist?
- What happens to our data if we cancel?
- What integrations are native versus custom?
- How do you handle outages and model changes that affect our workflow?
- Who do we call when it breaks?
When building makes sense
Build when the workflow is a source of competitive advantage, no product fits and you can staff it. In that case, build the smallest useful version first, document it and plan its maintenance from day one. Most owner-led companies in the 10 to 75 employee range will do better buying and configuring.
When hiring help makes sense
Bring in outside help when the decision is bigger than the tool: when you are unsure where to start, when nobody on staff has time to run a pilot or when you need someone to say no to the wrong project. Judge any provider by what they leave behind: a written description of what was built, a plan for who owns it and a way to measure whether it works.
Sources
- Fortune, summary of MIT NANDA "The GenAI Divide: State of AI in Business 2025"
- Gartner press release, June 25, 2025
Figures are quoted from the sources above and were reviewed in October 2026. Where a source is a secondary summary of other research, we say so in the text. Survey results reflect each survey's own sample and definitions. Examples labeled as illustrations use made-up numbers to show the method. Nothing here is a forecast or a guarantee of results.
Want this worked out for your business?
Tell us what is wasting your time.