The wrong way to Adopt AI

Welcome to The Logical Box!

For leaders who want AI to help, not add more work.

Hey,

If this is your first time here, welcome! The Logical Box is a weekly newsletter for owners and leaders who want AI to reduce real work, not add new work. Each issue focuses on one idea: see where work breaks down, fix the clarity first, then add AI where it actually helps.

If your business only works because you are in the middle of everything, this newsletter helps you build systems so it does not have to.

Now, on to this week.

A director at a mid-sized operations firm told me something last week that I keep thinking about.

They had just wrapped their quarterly review. The executive team spent 15 minutes celebrating their "AI transformation." They listed the tools: ChatGPT Enterprise, Copilot licenses for everyone, a custom GPT trained on their knowledge base, and an AI writing assistant for the marketing team.

Then they spent 45 minutes talking about why proposals still take 3 weeks to finalize, why client handoffs break, and why new hires ask the same questions veterans answered 6 months ago.

Nobody connected the two conversations.

The signal

If your company is racing to "adopt AI" right now, you are not alone.

According to Microsoft's latest AI diffusion data, roughly 1 in 6 people worldwide now uses generative AI tools. Adoption grew by 36 to 38% annually in 2025. Over three-quarters of organizations report using AI in at least one business function.

The pressure is real. Leaders feel it. Teams feel it. Everyone is moving.

But here is what nobody is saying in the LinkedIn posts and earnings calls.

The MIT Media Lab found that 95% of organizations see no measurable returns from their generative AI efforts. Gartner predicts that through 2026, organizations will abandon 60% of AI projects because the projects were built on data and workflows that were never ready for AI in the first place.

Only about 1 in 4 AI initiatives actually deliver their expected ROI. And when you look closer at who is getting results, fewer than one percent of organizations have mature AI deployments that deliver real, sustained value.

The gap between "using AI" and "getting value from AI" is massive.

And it is widening.

The pattern

Leaders are approaching AI adoption the same way: by starting with the tool.

They pick a platform. They roll out licenses. They run a training session on prompt writing. Maybe they appoint an AI champion. Then they wait for productivity to show up.

What happens instead is confusion.

Teams try AI on scattered tasks. Some people love it. Some people ignore it. A few create something useful. Most produce work that still needs to be rewritten, reviewed, or redone.

The tools are capable. The people are smart. But the work underneath is still unclear.

What does "done" mean on this proposal? Who owns the decision when a client asks for changes? What information does the AI actually need to write something the team can use without heavy editing?

Nobody clarified that before the tool showed up.

And AI does not fix unclear work. It multiplies it.

When a request is vague, the output is vague. When standards live in someone's head, AI cannot learn them. When handoffs break because context gets lost, adding AI just makes the breakdown faster.

This is the adoption pattern that is failing at scale right now. It treats AI like a software upgrade when it is actually a workflow redesign.

What is really breaking

The research on why AI adoption stalls points to three specific mistakes. These are not edge cases. These are the default paths most organizations are taking right now.

Mistake 1: They build horizontal platforms before proving vertical wins

Organizations see AI as an enterprise-wide capability, so they try to roll it out everywhere at once. AI for sales. AI for support. AI for HR. AI for operations. All running in parallel.

The instinct makes sense. If AI is transformative, why limit it to one team?

But according to enterprise AI research published in early 2025, this is one of the core failure modes. When you spread AI horizontally before you prove it works in one real workflow, you dilute the impact. You create coordination complexity. Accountability gets fuzzy. Nobody owns the result.

The organizations that are succeeding do the opposite. They pick one workflow. They rebuild it so it is clear and followable. Then they add AI where it removes repeat work. They prove value in 30 to 90 days. Then they copy that pattern to the next workflow.

Vertical first. Horizontal second.

Mistake 2: They stay stuck in the pilot phase

A team builds something promising. An AI chatbot for internal support. A document summarizer. A proposal generator.

It works in the demo. Leadership is excited. Then it goes live and nobody uses it consistently.

Why?

Because the pilot was built around the tool, not the workflow. The team did not integrate it into the places where people already work. They did not train people on how it fits their actual process. They did not change the incentives or expectations that would make adoption stick.

So the tool becomes optional. And optional tools do not get adopted at scale.

Industry research shows that fewer than 20 percent of AI initiatives have been fully scaled across the enterprise. Most organizations have pilots. Very few have systems.

The gap is not technical. It is operational.

Mistake 3: They treat AI as automation instead of workflow redesign

This is the biggest one.

Leaders hear "AI will make us more efficient" and they think that means speeding up what already exists. Automate the email. Automate the report. Automate the summary.

But if the underlying process is messy, automation just makes the mess faster.

A recent analysis of enterprise AI transformation efforts across multiple sectors reached the same conclusion: AI fails when leaders treat it as automation or a technology rollout. It succeeds when they treat it as a capability change, a mindset shift, and a workflow redesign.

That means before you add AI, you map the work. You name what is unclear. You define what "done" means. You assign ownership. You remove unnecessary handoffs. You standardize the inputs.

Then you add AI to handle the repeat tasks inside that now-clear system.

The sequence matters.

What works instead

The organizations that are actually getting ROI from AI share a pattern. It is not about the size of their budget or which models they use.

It is about how they sequence the work.

Step 1: Pick one workflow that repeats and hurts

Do not try to transform everything. Pick one workflow where the team feels the drag every single week.

Proposal creation. Client onboarding. Support ticket triage. Reporting. Quoting.

Make sure it repeats. Make sure it matters. Make sure the person who owns it wants it fixed.

Step 2: Make the workflow clear before adding AI

Map it end to end. Every step. Every handoff. Every input.

Then ask the clarity questions:

What does "done" look like here? Who owns each decision? What information is required before the work starts? What is the simplest version that still solves the problem? Where does work stall or get reworked?

Document the answers. Build a simple system the team can follow. Test it without AI first.

If the workflow does not work cleanly without AI, it will not work with AI.

Step 3: Add AI where it removes repeat thinking

Now you know where the system is solid. Add AI to the parts that repeat.

Generate the first draft. Summarize the client history. Pull the relevant data. Format the output. Check for common errors.

But keep humans in the decision points. Keep humans in the places where judgment and context matter.

AI is not replacing the work. It is removing the repetitive parts so the team can focus on what actually requires a person.

Step 4: Prove it in 30 to 90 days

Track two or three metrics that matter. Time saved. Cycle time. Error rate. Handoff breaks.

Do not wait six months to know if it worked. Get feedback fast. Adjust. Improve.

Research shows that companies with well-defined AI strategies and clear metrics within 30 days report 90 percent success rates in adoption and implementation.

The ones that skip this step stay stuck in endless experimentation.

Step 5: Copy the pattern to the next workflow

Once you have one win, you have a repeatable method.

You know how to diagnose unclear work. You know how to rebuild it into a system. You know where AI fits and where it does not.

Now you apply that same process to the next workflow. Then the next.

This is how AI scales. One proven system at a time.

The cost of getting it wrong

The data on failed AI adoption is not just about wasted software spend.

According to Boston Consulting Group research, 74 percent of companies struggle to achieve and scale the value of their AI initiatives. The organizations that get it right are pulling ahead. They are reinvesting their wins. They are building capability faster.

The gap between AI leaders and laggards is widening in 2026.

If you get the sequencing wrong, you do not just waste budget. You lose time. You lose credibility with your team. You train people to ignore the next initiative because this one did not deliver.

And while you are running pilots that never scale, your competitors are building systems that work without them.

Your next move to grow

Here is what you can do this week.

Pick one workflow that your team complains about. The one that always takes longer than it should. The one where people ask you the same questions. The one that breaks when someone is out.

Do not add AI to it yet.

Instead, write down three things:

1. What does "done" actually mean for this workflow?

2. Who owns the decision at each step?

3. What inputs are required before the work can start?

If you cannot answer those three questions clearly, your team cannot either. And AI will not fix that.

Start there.

Make the work clear. Then decide if AI can help.

A final thought

AI adoption is accelerating. The tools are improving. The pressure to move is real.

But the organizations that win in 2026 will not be the ones that adopted the most tools.

They will be the ones that fixed their workflows first.

If you want support while you build this, AI Clarity Hub is where I teach the process step by step. Monthly live sessions, tools, templates, and a group of leaders working through the same challenges.

But whether you join or not, the principle stays the same.

Fix the work. Then add AI.

Build your skills in here

I launched AI Clarity Hub, a private space for owners and leaders who want AI to reduce real work, not add new work.

Inside the hub, we work through exactly what I described today: finding the clarity gaps, building simple systems, and only then adding AI where it actually helps. Members get access to live training sessions, ready to use templates, and a library of AI assistants built for real business workflows.

Thanks for reading,

Andrew Keener
Founder of Keen Alliance & Your Guide at The Logical Box