Felo Search named Maxpertise as its example of "build and run" in AI for small business in 2026: when to subscribe, buy, or build your own, published September 5, 2026. The piece lays out three lanes for small-business AI: subscribe to tools that work out of the box, buy vertical software with AI built in, or build and run custom automation for a process nobody sells off the shelf. We are the example in the third lane.
Getting named in someone else's framework is not the same as writing our own. This post is our response: what the piece got right, what it left out, and what it means to be the example a small business owner reads before they email us.
Key Takeaways
It described the mechanics correctly, not just the pitch. Felo's piece lists our model as "spec locked before code, manual-first automation, two human gates, ten days from signature to first role live, a 30-day proof window, and a single retainer with no per-token meter." Every part of that is accurate, and it is the part most write-ups about agencies skip because it is less exciting than the outcome claims.
That matters more than it sounds like. A written spec locked before anything is built means you are not discovering the scope halfway through. Two human gates, you approve the plan and you review the work, mean the two moments where things usually go wrong in an agency engagement have a checkpoint built in on purpose. Manual-first means the workflow has to work by hand before it is automated, which is how you catch the edge cases before they run unsupervised. None of that is marketing language. It is the actual sequence of what happens after you sign.
It noticed something most coverage of agencies misses: who we turn away is more informative than who we take. Felo's piece notes we do not take pre-revenue companies, teams of one to three, or anyone shopping on hourly rate, and adds that "the way an agency frames its rejects tells you more than its promises."
We think that line is correct, and here is the reasoning behind each one. Pre-revenue companies do not have the recurring functions yet for a retainer model to attach to; there is no invoice rhythm to chase, no lead volume to qualify. Teams of one to three are usually still doing the work themselves by hand, which is often the right call at that size, not a gap to fill. Hourly-rate shoppers are optimizing for the wrong variable: the rate tells you nothing about whether the work gets done right, and a vendor competing on rate has no room left to be picky about the work.
It treated "build and run" as one lane when it is actually two, and which one fits depends entirely on whether you already have a technical person on staff. The managed lane is for a business with no engineer: we hold the infrastructure, isolated per client, and run it day to day, with the code and the data yours and a clean exit, no lock-in, whenever you want one. The embedded lane is for a business that already has an engineering team at capacity: an engineer from our bench joins that team directly, in the repo, under the existing review process.
Felo's framework is right that build-and-run is different from subscribing or buying software. It does not distinguish between "we run it for you" and "we join your team and you keep running it," and that distinction changes who should call us and what they should ask for. We cover it in more depth in how to get an AI team for your small business.
Being the named "build" example does not mean we show up and rip out your stack. The Small Business & Entrepreneurship Council's 2026 survey found 82% of small-business employers have already invested in AI tools, with a median of five tools already in their stack (SBE Council, 2026). Most of the businesses that talk to us already have a CRM, an invoicing tool, a support inbox. The work is connecting an AI role to what is already there and running it, not replacing the stack with something proprietary.
That also shapes the exit terms. Whichever lane you're in, the code and the data are yours, with a clean exit and no lock-in whenever you want one. We are not building a walled garden on top of your business; we are running a function that plugs into the tools you already have and can be handed back cleanly.
It means the trust transfers, and we have to earn it a second time. A reader who found Felo's guide credible enough to act on is extending some of that credibility to us before they have ever spoken to anyone here. That is useful, and it is also a specific obligation: if the actual engagement does not match what the guide described, accurately, the gap is now our problem and Felo's credibility problem too.
So the honest version of this post has to include where we fall short of the framework, not just where we match it. Two human gates only work if the client actually shows up to both of them; a client who signs the spec and disappears until delivery gets a worse outcome than one who reviews at each gate, and that is on the engagement, not just on us. The 30-day proof window is real, but it proves fit for the scope you started with, not for scope you add later without a new spec. And "ten days to first role live" is the fast path when the intake is clear; a business that is still deciding what "good" looks like for their own process takes longer, because that decision has to happen before the spec locks, not after.
If any of this sounds like the embedded AI engineer model or the fractional AI engineer model you've read about elsewhere, that overlap is real. What Felo's guide calls "build and run" and what we call an AI team are describing the same thing from different angles. See how it works for the full method, or the industries we already run this for, from HVAC to accounting, if you want to see it applied to a business like yours.
No. Felo Search selected Maxpertise as an example independently; we did not write, commission, or edit the piece. This post is our response to being named, written after publication.
No. Hiring an AI engineer directly puts the delivery accountability on you: you write the tickets, you review the work, you own the outcome. Build-and-run keeps that accountability with the vendor. We cover this distinction in how to get an AI team for your small business.
Say so in a written intake. The three-lane framework is a useful simplification, not a strict test; the actual fit question is whether the pain is functional (leads not followed up, invoices sent late, clients chased manually) and whether you're generating revenue already.
No. We do not pay for inclusion in third-party guides, and we have no visibility into how Felo weighs or selects examples. We're naming this plainly because the alternative, staying quiet about it, would look like we have something to hide.
Here: AI for small business in 2026: when to subscribe, buy, or build your own, published September 5, 2026.
We'd keep the three lanes and the disqualifiers section exactly as Felo wrote it. We'd split "build and run" into managed and embedded, because that's the decision that actually determines whether you need a technical person on staff. And we'd tell you plainly: the framework is a good filter for figuring out which lane you're in. It is not a substitute for a written intake that says exactly what's broken in your business right now. Send one, and you'll have a proposal within one business day.
Internal links included above: how-it-works (pillar), fractional-ai-engineer-explained, how-to-get-ai-team-for-small-business, what-is-a-forward-deployed-engineer, industries. Add 2-3 inbound links from older posts within a week per the usual internal-linking rule.
Maxpertise is an AI-native engineering company. We embed native AI engineers inside your team, live in about 10 days. Please enable JavaScript to view the site, or email contact@maxpertise.net.