Most businesses do not need another AI brainstorm.
They need help with a much simpler question:
What is the first place in this business where AI can create a clear, visible result without making everything messier?
This is the gap I keep coming back to.
AI adoption is moving fast, but the day-to-day reality is still messy. Plenty of businesses have tried ChatGPT, Copilot, a few automations, maybe a workshop or two. That part is not rare anymore. The hard part is making any of it fit the way the business already works.
The short version:
Traditional businesses do not need an "AI transformation program" as their first step. They need one working use case that saves time, reduces admin, or improves turnaround inside an existing workflow.
The shift that matters
McKinsey has a useful distinction between off-the-shelf adoption and what it calls Shaper use cases.
- Off-the-shelf adoption is when a company buys a general tool and uses it as-is.
Shaperwork is when a company connects prompts, data, and systems into a workflow that matters to the business.
I think that distinction explains why so many AI pilots stall.
Buying a tool is easy. Getting it to behave inside quoting, reporting, approvals, client communication, research, or operations is the real job. That is where the value is. It is also where the headaches start.
What businesses are actually struggling with
The survey data is pretty consistent on this.
RSM's 2025 middle-market survey says 70% of respondents need outside help to get the most out of their AI solutions. The detailed business and professional services breakdown shows why:
| Challenge | Share |
|---|---|
| Data privacy and security concerns | 40% |
| Data quality concerns | 40% |
| Difficulty integrating AI into existing workflows | 34% |
| Intellectual property concerns | 29% |
| Insufficient internal skills and expertise | 26% |
| Regulatory or compliance concerns | 26% |
And among firms that felt underprepared, the top reasons were:
| Reason they felt underprepared | Share |
|---|---|
| Lack of in-house expertise | 47% |
| Difficulty identifying the right AI use cases | 42% |
| Lack of clear AI strategy | 33% |
| Data quality challenges | 33% |
| Difficulty selecting the right AI technology | 33% |
That pattern matters.
Businesses are not mainly asking for help with model hype. They are asking for help with:
- choosing the right first use case
- connecting AI into real workflows
- improving trust in the underlying data
- reducing privacy and compliance risk
- proving that the investment is worth continuing
Why the first win matters
For a traditional business owner, the first win has to be concrete.
Not:
- "we have an AI roadmap"
- "we are experimenting with agents"
- "the team is more AI-enabled"
But:
- "quotes now go out the same day"
- "handoff notes are generated automatically"
- "client research takes 20 minutes instead of 2 hours"
- "staff can find the right answer without interrupting a manager"
That is the difference between AI as a talking point and AI as something people actually use on a Wednesday afternoon.
A better ladder for traditional businesses
1. AI Workflow Review
Find the best first use case and the main blockers.
Review the current tools, the messy handoffs, the obvious risks, and the fastest route to one useful result.
2. First Working AI Use Case
Make one important workflow actually work.
Build one real AI-assisted workflow in quoting, reporting, client follow-up, knowledge lookup, or admin.
3. Make AI Part of the Business
Turn the first win into a normal way of working.
Add documentation, ownership, safeguards, and rollout steps so the pilot stops feeling fragile.
4. Ongoing AI Support
Keep improving what works.
Refine the workflow, review how it is being used, tighten the guardrails, and add the next worthwhile use case.
The language problem
One reason AI service offers often miss the mark is that they are written in provider language rather than business-owner language.
Labels such as:
- agent enablement
- MCP integration
- prompt engineering
- multi-agent workflows
may mean something to the people building the system, but they are not how most business owners describe their problems.
A business owner is more likely to respond to:
- reduce admin time
- speed up turnaround
- connect the tools you already use
- make one workflow work properly
- show a clear before-and-after result
The technical layer still matters. A lot.
It just should not be doing the selling.
What the better firms are really selling
When you look closely, the better AI service firms are not really selling prompts or model access.
They are selling some version of this:
flowchart LR A["Scattered AI experiments"] --> B["Choose one meaningful use case"] B --> C["Connect AI to real tools and data"] C --> D["Prove a visible business result"] D --> E["Roll it into day-to-day operations"] E --> F["Improve and expand carefully"]
That is a much stronger position than "we do AI consulting." It is also easier for a buyer to understand.
What a sensible first conversation sounds like
It should sound more like this:
Where is your team losing time every week?
Which step gets delayed because information is hard to find, rewrite, or hand off?
If we made one of those steps faster and clearer in the next month, what would matter most?
That gets you much closer to value than opening with model names, orchestration diagrams, or futuristic language.
The real opportunity
The opportunity is not to give traditional businesses more AI noise.
The opportunity is to help them move from:
- experimentation to execution
- tool sprawl to workflow clarity
- vague promise to visible result
That is where trust starts.
It is also where repeat work comes from, because once one workflow is working, the next conversation gets a lot easier.
That is the version of AI that feels useful to me. Less theatre, more follow-through.