OMNINEYou have done the hard first part. The week of logging is over, the Friday tally is on paper, and you are holding what most owners never produce: a named problem with a number attached, the quote rebuilt by hand on every job, the booking that dies between the calendar and the confirmation, the hours a week your operation spends moving information instead of acting on it. Then comes the fork that decides whether the number turns into a fix or a regret. Every capability a business adds arrives by one of three roads: built for you, bought off the shelf, or connected out of pieces you already own. Nearly every wasted AI dollar I have seen was a wrong turn at this fork, taken confidently, so this piece lays the fork out as a decision tree, with the root questions that come before the branches, the anchored questions that test each branch, the failure mode each branch produces when chosen wrongly, and the reason the connect branch quietly wins most small business decisions.
The root questions
Before any branch, two questions route the whole decision, and skipping them is how owners end up comparing vendors for a problem no vendor has.
The first root question: does the capability already exist in pieces you own? Most operations are carrying more software than they use, subscriptions accumulated one urgent Tuesday at a time, and an honest inventory of the current stack regularly turns up the scheduler, the form tool, or the automation feature the new purchase was about to duplicate. The accumulation pattern, and why it makes businesses slower rather than faster, is the subject of why buying more AI tools is making you slower. Take the inventory before the catalog, every time.
The second root question: is this a capability problem or a flow problem? A capability problem means the business cannot do the thing at all, cannot transcribe calls, cannot generate drafts, cannot forecast demand. A flow problem means the business can do every individual step and the steps do not connect, so a person carries information between them by hand. The distinction matters because the two problems route differently. Capability problems continue down the tree to build or buy. Flow problems route to connect by default, because the failure lives between systems that each work fine alone, and my research is part of why I trust that default: the leaders who adopted successfully were solving recognized communication breakdowns, information fragmented across disconnected channels, which are flow problems wearing different clothes (Goodwin, 2014).
The build branch
Building means commissioning or writing software that did not exist before, and it is the right road in a narrow set of conditions: the capability is your product or a real differentiator, nothing on the market fits after an honest search, and you can own the maintenance for years, not months. Three anchored questions test the branch. Will this create an advantage a competitor cannot simply buy next quarter? Who maintains it in month thirteen, named as a person rather than a hope? And what breaks when that person leaves? If the answers are no, nobody in particular, and everything, the branch is wrong however exciting the mockup looks.
The failure mode of a wrong turn here is the most expensive on the tree, because a build is the only branch where the cost keeps arriving after the invoice. Custom software is a commitment to updates, fixes, and compatibility work for as long as it runs, and a small business that builds what it could have bought has purchased a second, invisible product: the permanent job of keeping the first one alive. The cost shape is the highest fixed outlay on the tree plus an ownership meter that never stops, and for most owner operated businesses, most of the time, that shape alone disqualifies the branch.
The buy branch
Buying means adopting a finished product, and it is the right road when the capability is a commodity, the category is mature, and the tool replaces an entire workflow rather than adding a surface to an existing one. The anchored questions here are about edges, because a purchased tool succeeds or fails at its edges. Does this eliminate movements of information end to end, or does it add a login where information can sit without being shared? What does it connect to in your existing stack on day one, demonstrated rather than promised? And what does leaving cost, in exported data and rebuilt habits, if the tool disappoints in a year? A leader’s full evaluation lens, aimed at value and fit rather than feature counts, is built in how to evaluate an AI tool as a leader, not an engineer.
The failure mode is accumulation, and the evidence on where it leads is blunt. BCG’s survey work found that the companies struggling to get value from AI spread themselves across six or more use cases per workflow while the leaders pursue about three and go deeper, and that only about a quarter of companies had moved beyond proofs of concept to real value at all, with a mere 4 percent capturing substantial value (BCG, 2024). Every additional purchase without an integration plan is another place information can live unshared, which means the buy branch, taken carelessly, manufactures the exact flow problems the tree was supposed to solve. The cost shape is a subscription that never ends plus integration debt that compounds, and the branch earns its keep only when the tool retires a workflow outright.
The connect branch
Connecting means making the pieces you already own exchange information without a human courier, and it is the right road whenever the root questions found a flow problem: the scheduler that does not tell the payment system, the intake form that does not reach the calendar, the quote tool that lost its supplier library. The anchored questions are diagnostic rather than commercial. Where exactly does the information stop flowing, traced as a path rather than guessed as a tool? What are the two systems keyed on at the break, names, codes, or identifiers, since mismatched keys are where connections quietly die? And which human is currently the bridge, because that person’s manual chore is the specification for the fix. One caution anchors the branch: connect only what a named problem says to connect, because wiring systems together ahead of a diagnosis just automates confusion. The stage that produces the named problem is opened up in inside Assess: what the first stage really measures.
The failure mode is rarer and cheaper than the other branches’ failures, which is part of the argument. The cost shape is mostly one time work with the smallest recurring footprint on the tree, and the value evidence points the same direction: McKinsey finds that redesigning how work flows end to end is the change most associated with bottom line impact from AI, well ahead of adding tools onto existing processes (McKinsey, 2025). Connection is workflow redesign at small business scale, done with the software already paid for.
Set the three cost shapes side by side and the tree gets easier to hold in your head. Build is the largest fixed outlay plus a maintenance meter that never stops running. Buy is a smaller entry price plus a subscription forever and integration debt accruing at the edges. Connect is mostly one time work applied to tools already paid for, with the smallest recurring footprint of the three. None of those shapes is wrong in itself, and the tree’s whole job is matching the shape to the problem in front of you rather than to whichever sales pitch arrived first.
Two decisions from the field
Advocate Roofing felt like a buy decision. A version migration had left their quoting platform generating estimates without materials, the team was rebuilding every quote by hand, and the obvious move was to replace software that seemed broken. The tree routed elsewhere. The root questions found a flow problem between two healthy systems, the diagnosis traced the break to a naming layer where template product names no longer matched the supplier library, plus a redundant supplier connection interfering underneath, and the repair took an afternoon of systematic work instead of a migration to a new platform that would have transplanted the same fault. Repair over repurchase, and the full story is in how a roofing company stopped rebuilding every quote by hand.
Lucky Little Riders, the youth riding school, felt like a different buy decision: the all in one platform that promises scheduling, payments, and client management in a single subscription. The tree routed to connect. The pieces already existed, a scheduling tool, a payment system, and an intake and waiver flow, and the problem lived in the gaps between them, so the work joined them into one continuous path where a parent books once and everything else follows without anyone rekeying anything. The first act of that build is told in the first AI workflow every service business should build, and in both cases the deliverable was the same thing the tree keeps pointing at: information the business already had, arriving where it was needed, without a person carrying it.
Why connect wins most small business calls
The pattern across Omnine’s field work is consistent enough to state as a default with reasons attached. Small businesses mostly do not have capability gaps. They have working tools that were never introduced to each other, which makes their expensive problems flow problems, and flow problems route to connect. Their scale rarely justifies the maintenance meter that build switches on. And their tool shelves are usually full already, which makes another buy more likely to add a surface than retire a workflow. The default has exceptions the tree surfaces on its own, a true capability gap routes to buy, a product idea routes to build, but as a starting posture, look between your systems before you look at a catalog. The everyday cost of not doing so, the tax a business pays whenever a person is the transport layer between its tools, is priced in the communication tax your business pays every day.
The pushback
Connect sounds like duct tape, and a platform would be cleaner. A platform is a buy, and a buy still has to meet your remaining tools, your data, and your team’s actual habits at its edges, which means the platform decision is a connection decision wearing better marketing. All in one is a promise about the inside of the box, and the failures documented across this blog happen at the edges of the box. Judge any platform by the buy branch’s edge questions and it either passes or it does not, but cleaner is not a property a subscription can grant by itself.
A custom build would future proof us. The month thirteen question is the answer, asked without sentiment. Software does not future proof a business; a maintained system does, and maintenance is the cost most build decisions forget to price. For the rare small business whose product is the software itself, the branch is right and the question has a real name in it. For everyone else, future proofing is better purchased through connections that can be rewired as tools change than through code that must be rewritten.
We have nothing worth connecting. The inventory usually disagrees, and the exercise costs an hour. But when the shelf truly is bare, the tree still has advice: buy small, buy for the edges, and connect from the first day, so that the second tool you ever own arrives already talking to the first. Businesses that start with the connection habit never accumulate the disconnected pile that this whole decision exists to manage. Where the tree sits inside the larger method, as the opening act of the Architect stage, is shown in how an Omnine engagement runs.
Run the tree this week
Take your named problem, the one with the number on it, and walk it down from the root. Answer the two root questions in writing, follow the branch they point to, and put that branch’s anchored questions to whatever solution you were already leaning toward. Most owners discover in under an hour either that their instinct survives the questions, which is worth knowing, or that a cheaper road was sitting in plain sight, which is worth more. And if you do not yet have a named problem to route, that is the tree telling you that you are one step early: the Omnine AI Readiness Assessment manufactures the starting point, an honest read of where your leverage is, in about three minutes.
References
Goodwin, M. R. (2014). A qualitative descriptive multiple-case study: Fortune 500 leaders’ social business platform adoption (Doctoral dissertation, University of Phoenix). ProQuest Dissertations Publishing (UMI No. 3648813).
McKinsey & Company. (2025). The state of AI in 2025: Agents, innovation, and transformation. McKinsey Global Survey on the state of AI.
Boston Consulting Group. (2024). Where’s the value in AI?