Does your business need AI? It is the wrong first question, and that is why so many mid-market operators waste money answering it. The honest answer for most firms between 2 million and 30 million in revenue is no, you do not need a custom model. You need closed loops on the processes you already run. This is a readiness test, not a sales pitch. By the end you will have a score and a clear next step, even when that step is to wait.
We run an audit-first practice. We look at where your business leaks revenue, rank the leaks by dollar impact, then build and run the fix only if the math holds. Sometimes the math says wait. So let us start where most advice refuses to.
Why most mid-market businesses do not need a custom AI model
The market sells custom models because they are expensive and impressive. Reality is duller. A custom model solves problems most operators do not have. Your quote follow-up does not need a fine-tuned transformer. It needs to happen every time instead of half the time. That is a closed loop, and a closed loop usually beats a clever model.
The failure data backs this up. Gartner predicted that at least 30 percent of generative AI projects would be abandoned after proof of concept by the end of 2025, citing poor data quality, weak risk controls, and unclear business value. An MIT report found that 95 percent of generative AI pilots delivered little or no measurable impact on the bottom line. The pattern is not bad technology. It is good technology aimed at the wrong target.
So before you ask whether you need AI, ask what you are actually trying to fix. We wrote more on this trap in most AI implementation is theater. The short version: a demo that impresses your board is not a system that moves your numbers.
What does AI readiness actually mean?
Readiness is not about how modern you feel. It is about four boring conditions being true at once. You have repeatable high-volume work. Your data is clean enough to act on. You can name a specific revenue leak. And your team has the capacity to adopt a change. Miss any one and your project joins the abandoned pile.
That capacity point matters more than people expect. BCG found that only about 5 percent of companies were generating substantial value from AI, with the gap driven less by tools and more by how people and processes absorbed them. The tool was rarely the bottleneck.
If you want the formal version of the four conditions, read our closed-loop score framework. It is the scoring spine behind the test below.
Signal 1: process maturity
Can you write down the steps of the process you want to improve? If a new hire could follow the steps without asking three questions, the process is mature enough to automate. If it lives in one person's head, fix that first. Automating chaos just produces faster chaos. This is the most common reason we tell operators to wait.
Signal 2: data hygiene
You do not need a data lake. You need data that is consistent enough to trust. One source of truth for customers. Fields filled in the same way. Gartner has warned that a lack of AI-ready data puts most projects at risk, predicting that 60 percent of AI initiatives unsupported by ready data would be abandoned. Messy data is a no, not a maybe.
Signal 3: volume threshold
Automation pays when it runs often. A task you do twice a year is not worth a system. A task you do 50 times a week is. If a process touches more than roughly 100 events a month and each event carries a clear dollar value, it clears the volume bar. Below that, a checklist beats a build, and the cheaper fix is usually the right one.
Signal 4: team capacity to adopt
This is the signal most operators skip, and it sinks more projects than any technical gap. A system someone has to learn, monitor, and trust needs an owner with time to do that. If your best person is already at 110 percent, a new tool becomes shelfware within a month. Capacity is not a nice-to-have. It is the difference between a system that runs and a license you pay for and forget. Free up the owner first, or the build will fail no matter how good it is.
Why is most AI value back-office, not front-office?
Here is a finding that surprises most operators. More than half of corporate AI budgets go to sales and marketing tools, yet the MIT research found the biggest measurable return sat in the back office, cutting manual operations, reducing outside agency costs, and removing handoffs nobody enjoyed. The flashy front-office demo wins the meeting. The dull back-office loop wins the quarter. When you run your own readiness test, weight the boring processes higher than the exciting ones. Invoice matching, data entry between systems, and overnight ticket triage rarely make a highlight reel, but they leak real money every single day, and they are the easiest loops to close cleanly.
Where are the leaks, and do they justify a build?
Readiness without a leak is just curiosity. The leaks we see most often in mid-market firms are slow lead response, quotes that never get followed up, manual data entry between systems, and support tickets that pile up overnight. Each one has a number attached. Late lead response alone can cut conversion by half, which is real money you can measure this quarter.
We named the most expensive of these the open loop tax. It is the revenue that quietly drains out of unfinished processes. Read the open loop tax to see how it adds up, or run the numbers yourself with the open loop tax calculator. If you are unsure which process to start with, our guide on where to start automating operations in mid-market walks through the triage.
The point of finding the leak is honesty. If your biggest leak is worth 2,000 a year, no AI build is worth it. If it is worth 200,000 a year, you have a case. The dollar figure decides, not the excitement. We rank every leak by annual dollar impact and ignore anything that cannot be measured, because a leak you cannot size is a leak you cannot prioritize. This is also why a vague yes to the AI question is dangerous. Yes to what, fixing which leak, worth how much. Without those answers you are buying motion, not results, and motion is exactly what fills the abandoned-project pile.
The mid-market AI readiness test, scored
Score one point for each true statement. Be strict. Optimism here costs you later.
1. You can document the target process end to end. 2. Your customer data lives in one trusted system. 3. The process runs at least 100 times a month. 4. You can name one leak and attach a dollar figure to it. 5. At least one person has time to own the rollout. 6. Leadership agrees the leak is a priority this quarter.
Five or six points means you are ready to build. Three or four means fix the process or data first, then revisit. Two or fewer means you are not ready yet, and any vendor who tells you otherwise is selling, not advising. You can also run our operator scorecard for a faster read on the same signals.
Ready now, fix-first, or not yet?
The table below maps the signals to the three honest outcomes. Find the column that matches your reality.
| Signal | Ready now | Fix process first | Not yet |
|---|---|---|---|
| Process documentation | Written and repeatable | Partly written | Lives in one head |
| Data hygiene | One trusted source | Cleanable in weeks | Scattered and conflicting |
| Volume | Over 100 events a month | Growing toward it | Rare or seasonal |
| Defined leak | Named with a dollar figure | Suspected, not measured | Unknown |
| Team capacity | An owner with time | Owner needs freeing up | No one available |
Most mid-market operators land in the middle column on their first read. That is good news. Fix-first is cheap and fast, and it sets up a build that actually works instead of one that gets abandoned.
When does a custom build make sense?
Rarely, but not never. A custom build earns its cost when the work is high volume, the leak is large, and no off-the-shelf tool fits your specific workflow. We laid out the exact threshold in when to build a custom AI app. For everything below that line, buying or wiring existing tools wins. The trade-off is covered in build vs buy software for mid-market.
The market still scales fast on paper. McKinsey reported that 88 percent of organizations used AI in at least one function, yet only about a third had moved past piloting. Adoption is not value. The firms that win pick one painful, expensive process and close the loop, then move to the next.
What does the free audit actually tell you?
The readiness test above gives you a strong self-estimate. The audit gives you the real number. We map your processes, rank the leaks by dollar impact, and tell you which ones are worth fixing and which are not. If nothing clears the bar, we say so. That honesty is the point, and we explain why in why kratt gives the free AI audit away.
If you are weighing who to trust with this, the difference between an audit-first practice and a typical vendor is structural, not cosmetic. We unpack it in AI consultancy vs agency. When the audit confirms a fix is worth building, our automation service builds, hosts, and runs it, so the loop stays closed without adding headcount.
Common questions about AI readiness
Does your business need AI if revenue is growing fine?
Growth hides leaks, it does not close them. A business growing at 20 percent can still lose 15 percent of leads to slow response. The readiness test ignores how you feel about growth and looks at whether a measurable leak exists. If it does, the math is the math.
Is my business ready for AI without a data team?
Often yes. You do not need a data team to clean one customer database or document one process. The readiness conditions are operational, not technical. If your data is consistent and your process is written down, you are further along than most firms with full data teams.
How do I know if I should fix the process before adopting AI?
If you scored three or four on the test, fix first. The clearest tell is whether a new hire could run the process from your written steps. If not, the process is the bottleneck, and no system will rescue an undefined workflow.
What are the clearest signs you need AI automation?
Repeated manual handoffs between systems, work that piles up overnight, and follow-ups that depend on someone remembering. These are the signs you need AI automation, because they are high volume, low judgment, and expensive when missed. They are also the easiest loops to close.
When is the right time to adopt AI in mid-market?
The right time to adopt AI in mid-market is after a specific leak is measured and a process is documented, not before. Timing is about readiness conditions, not market pressure. Adopting because competitors did is how firms end up in the 95 percent that see no return.
Will the audit just tell me to buy your services?
No. Roughly a third of the audits we run end with a recommendation to fix a process or wait, not to build. We only build when the dollar impact justifies it. An audit that always says yes is a sales call wearing a lab coat.
So, does your business need AI? Run the test, find your column, and then get the real number. Take the free AI audit and we will rank your revenue leaks by dollar impact, then tell you honestly whether a fix is worth building. Every system we ship carries our Recovery Guarantee: if it does not recover the revenue we projected, we keep working until it does, at no extra cost.