98% of Companies Are Implementing AI. Only 5% See Results.
98% of Companies Are Implementing AI. Only 5% See Results.
In February 2026, McKinsey’s Joe Ngai shared a number that should make anyone investing in AI sit up: 98% of executives say they’re implementing AI, yet only 5% see bottom-line impact [1].
MIT’s NANDA research tells the same story: 95% of GenAI pilots fail to deliver measurable P&L impact [2].
The technology works. The tools are accessible. So what’s going wrong?
Why most AI implementations don’t deliver ROI
Most AI usage in SMEs looks like this:
- Faster emails
- Better first drafts
- Quick summaries
- “Can you rewrite this?”
Useful? Yes. Visible on the P&L? Almost never.
There’s a fundamental difference between an individual shortcut (scattered, hard to measure) and a process change (repeatable, measurable, scalable).
The 95% that fail are mostly doing the first. The 5% that profit are doing the second.
The real problem: horizontal AI vs. vertical AI implementation
Most business owners don’t separate generic AI from specific solutions. To them, it’s one thing: “AI.”
That’s understandable — they’re running a business, not following model benchmarks. But it creates a practical problem:
- Horizontal AI (like a general copilot) spreads thin gains across everything.
- Vertical AI (built around a specific workflow) concentrates big gains on one thing.
MIT found that specialized vendor solutions succeed about 67% of the time, while internal builds succeed about 33% [2]. The gap comes down to specificity. The more tightly AI integrates into an actual workflow, the more likely it delivers.
The companies that get this right tend to look boring from the outside. They’re not chasing the flashiest model or the most autonomous agent. They’re automating invoice processing. Flagging website issues. Drafting proposals. Checking SEO rankings. Reviewing contracts. Posting content on schedule.
Goldman Sachs and AT&T deployed their first real AI agents for compliance, onboarding, and fraud detection [3]. The boring stuff. The repetitive stuff. That’s where the ROI lives — in faster responses and fewer errors.
SMEs have an advantage (most don’t realize it)
There’s a popular narrative that companies need to “redesign the entire organization” to succeed with AI. That’s a useful framework for large enterprises with 12 layers of management.
For SMEs, it’s different. And honestly, easier.
In most SMEs, the owner is also the line manager. They know where the money goes, where time is wasted, and where the bottlenecks live. There’s no politics, no innovation committee, no 18-month transformation roadmap.
You don’t need to redesign the org. You need to rethink specific processes — with a clear method and someone to guide the implementation.
How to implement AI in a small business: start with an internal assistant
When someone asks me where to begin, I almost never start with complex automations.
I start with an internal AI assistant. Think of it as hiring an intern — one that’s available 24/7, eager to learn, and gets better with every task. Over time, through consistent delivery and growing trust, that intern works its way up. You start with small tasks, expand as confidence builds, and gradually hand over more responsibility.
The assistant can help with things you’d never think to automate individually:
- Identify issues on your website and check your SEO rankings
- Propose and write blog posts based on your expertise
- Create social media content and schedule it consistently
- Help prepare for meetings and draft proposals afterward
- Review legal documents and flag what matters
- Manage invoices and follow up on late payments
- Manage your calendar and send reminders to clients
- Answer recurring internal questions so people stop interrupting each other
None of this is glamorous. All of it saves hours every week — and more importantly, it reduces errors. A system that drafts proposals doesn’t forget your pricing. A system that posts content doesn’t skip a week because someone got busy. A system that reviews contracts doesn’t skim the last three pages on a Friday afternoon.
Treat it like a real team member
One practical principle I’ve found essential: give your AI assistant its own accounts and credentials, just like you would any other employee. Its own email, its own user accounts, its own API keys.
Why? The same reasons you don’t share passwords between employees — access control and auditability. When the assistant has dedicated credentials, you control exactly what it can see and do. You can review what it accessed, revoke permissions if needed, and share with it only what the task requires. This also makes compliance straightforward, because every action is traceable to a specific identity.
The pattern: discover first, then automate
Once the assistant finds a repeatable pattern, you turn that pattern into a rigid process — more efficient, auditable, and consistent.
A proposal system might start as something manual: templates, text snippets, rules in someone’s head. The healthy path:
- The assistant helps generate consistent drafts.
- You identify the format that actually works.
- Only then do you automate: data in, document out, with version control and quality checks.
That’s how a proposal generator I built evolved — from a custom script into a generic tool that converts structured content into branded PDFs. The assistant discovers the process; the automation locks it in.
A practical framework: 4 areas where small businesses see real AI ROI
Instead of “an AI project” (vague), I work more like a personal trainer: monthly, ongoing, focused on results, adjusting as the business learns.
Almost everything I do falls into one of four areas — and you can start with just one:
1) Internal AI assistants
An assistant that knows the company: products, pricing, policies, documentation, internal FAQs. Goal: cut time spent searching, re-explaining, and context-switching.
2) Workflows and automations
Reports that build themselves, approvals that follow a clear path, alerts when something’s off. Goal: remove repetitive tasks from the team’s plate.
3) Browser automation (when there are no integrations)
Some business software doesn’t talk to anything. A robot does the clicks: extract, fill, validate. Goal: scale without hiring for keyboard tasks.
4) Customer-facing chatbots
Not to look modern — to answer the basics, capture leads, qualify requests, and free up the team. Goal: better response times and no missed opportunities outside business hours.
The right question is: which of these four areas takes the most pressure off your business in the next 30 days?
Want to close the gap? Let’s talk.
If your team is already using AI day-to-day but you can’t point to a clear impact — in time, cost, quality, or revenue — I can help you pick one process to tackle first and turn it into a repeatable system, with monthly support.
Send me a message and we’ll have a short conversation to figure out where it makes most sense to start.