Webinar: Why Do AI Adoption Projects Fail (And How to Make AI Actually Work)?
Kyndall Elliott, Marketing Lead at Workzone, sat down with Brian Kohlmann, Director of AI Emerging Technology and Martech at Bader Rutter, to talk about what actually happens when businesses try to adopt AI. Here’s what they got into.
Short version: Most failed AI rollouts aren’t a technology problem. They’re a process and data problem. AI amplifies whatever workflow you point it at, so teams that automate before fixing their process just get broken output faster. The teams that win diagnose the real problem first, pick low-stakes use cases, vet vendors with hard tests, put governance in before scaling, and build the process before they automate it.
Every vendor deck right now promises the same thing: buy the AI, and your problems disappear. Your team moves faster. Your backlog clears. Your headcount stretches twice as far.
Then you actually roll it out, and month three looks a lot like month one. The tool is powerful. Nothing got easier. And nobody can quite explain why.
Kyndall has lived this as a marketer leaning hard on AI to get more done. Brian has watched it play out across dozens of companies. They landed on the same conclusion from opposite ends: AI is a good tool aimed at the wrong question. Most teams ask “how do we add AI?” when the question that matters is “what’s actually broken, and is AI the thing that fixes it?”
Here’s what came out of the conversation.
Watch the full session here:
Should you adopt AI to solve a business problem?
Not before you know what kind of problem you have. AI is a tool, not a decision, and pointing it at the wrong problem wastes money and time.
Brian’s framing is blunt. “AI is a powerful tool, but it’s not a panacea. It’s not a one size fits all for every application.”
Kyndall put the flip side just as plainly: “AI is great in a lot of ways, but it can also be dangerous in a lot of ways.”
The danger isn’t the technology. It’s skipping the diagnosis. Before you reach for an AI solution, figure out whether the problem you’re solving is a people problem, a process problem, or a technology problem. A lot of the time it’s one of the first two, and no model is going to fix a workflow that was never built. Start with the problem, not the tool. Have the unglamorous conversation about what’s actually going wrong before you shop for something to throw at it.
What are the best use cases for AI in business?
The best AI use cases are low-complexity, repetitive tasks: synthesizing documentation, transcription, first-pass competitive analysis, and building briefs and templates. Accuracy and governance risk climb as complexity climbs, so the deeper AI runs into critical workflows, the more oversight it needs.
Brian sorts AI use cases into two buckets: helpful, and not. The helpful bucket is the repetitive work that eats an afternoon and produces something a person still reviews. Some of these wins don’t even need AI. He’s automated entire workflows in Python that cut hours down to seconds, no model required.
The trouble starts as complexity climbs. “The low hanging fruit of synthesizing documentation, doing competitive analysis, helping you build out briefs, those are great use cases. But the more complexity you build into that, the harder it becomes to maintain accuracy and governance.”
Deep, autonomous integration into critical workflows is where the rewards get bigger and the failure modes get expensive. Accuracy slips. Nobody’s accountable. A confident wrong answer ships to a client. Match the tool to the task. Let AI take the repetitive, low-stakes work, and keep a human between the model and anything that carries real consequences.
How do you evaluate an AI vendor or tool?
Run five tests before you buy: the pricing page test, the turn-it-off test, the roadmap test, the failure-mode test, and the governance test. A tool that can’t pass them will not get better once it’s inside your stack.
Both Kyndall and Brian have felt the pressure from the top. “When your executive team says, ‘If we get AI, it’ll solve all our problems,’ it’s tempting,” Kyndall said. “But without diligence, that’s a recipe for trouble.”
Here are Brian’s five tests:
- The pricing page test. Does AI look bolted on, or built in? Sloppy integration on the pricing page usually means sloppy integration in the product.
- The turn-it-off test. Can the tool still do its job with the AI switched off? If not, you’re buying a demo, not a product.
- The roadmap test. Is the AI roadmap deliberate, or is it a feature list chasing headlines?
- The failure-mode test. When the system gets it wrong, who is accountable? If the answer is “nobody,” that’s your answer.
- The governance test. Are security, data retention, and governance protocols actually spelled out, or hand-waved?
Run these before you sign anything.
Why does AI governance matter?
Because AI moving fast is only good news if someone can see where it’s going. Without oversight and clear accountability, AI investments get made erroneously and mistakes ship without anyone catching them.
Governance sounds like the boring part. It’s the part that decides whether AI helps you or quietly burns money. Brian doesn’t soften it: “If there’s nobody watching the watchers, investments are made erroneously or dangerously.”
Kyndall framed it as a question every buyer should sit with: “If you’re not doing the due diligence, what are you signing up for on the back end that might get you in trouble later?”
Assign accountability. Build oversight into the decision, not after it. Make the risks and the benefits visible to the people signing off.
How do you implement AI in your workflow the right way?
Build the process first, then bring in AI to run it. AI amplifies whatever it’s pointed at, so a clean workflow gets faster and a broken one produces broken output faster. Onboard AI models the way you’d onboard a new hire, with clear guidelines and context.
Kyndall’s own rollout is the cautionary tale. “When I started implementing AI in my workflow, I didn’t have a process. It made everything take longer. AI doesn’t create the process for you.”
That’s the whole lesson in two sentences. Brian’s mental model helps here: treat AI models like new employees. “You have to onboard AI models the same way. Give them guidelines and proper context.” You wouldn’t hand a new hire zero direction and expect great work. Careless copy-and-paste with no context in and no review out is how output quality quietly degrades.
Can AI fix broken processes or bad data?
No. AI can’t paper over bad data, tangled processes, or the wrong people in the wrong seats. Used on a weak foundation, it makes existing problems worse. As Brian puts it, there is no AI strategy without a good data strategy.
“There is no AI strategy without a good data strategy.” It’s the line to tape to the wall. Kyndall watched the alternative play out firsthand when she adopted too early: “All that did was make everything take twice as long.”
Fix the foundation before you automate on top of it. AI does its best work in structured, well-run environments. Drop it into a mess and all you’ve done is speed up the mess.
The takeaway
Kyndall and Brian didn’t leave with a pitch for AI or a warning against it. They left with a sequence. Diagnose the actual problem. Pick the use cases where AI genuinely earns its place. Vet your vendors with real tests. Put governance and accountability in before you scale. Build the process before you automate it. Get your data house in order first.
Brian’s close said it cleanly: “If you treat AI like any other tool, with respect, governance, and a clear purpose, it’ll deliver far more value and credibility to your organization.”
Whether you’re a solo marketer figuring out where AI fits or an enterprise weighing a full rollout, the work is the same. AI is not the strategy. It’s what you use once the strategy is sound.
Frequently asked questions
Why do AI adoption projects fail?
Most AI adoption projects fail because teams treat AI as a solution to a problem they haven’t diagnosed. AI amplifies the workflow it’s pointed at, so pointing it at a broken process produces broken output faster. The failure is usually a process or data problem, not a technology one.
What business problems is AI actually good at solving?
AI is strongest on low-complexity, repetitive tasks: synthesizing documentation, transcription, first-pass competitive analysis, and drafting briefs and templates. These are low-stakes, high-volume jobs where a human still reviews the output. Risk to accuracy and governance rises as you push AI deeper into critical, autonomous workflows.
How do you evaluate an AI tool before buying it?
Run five checks. Does the pricing page suggest AI is built in or bolted on? Can the tool function with AI turned off? Is the AI roadmap deliberate? Who is accountable when the system fails? Are security, data retention, and governance protocols clearly documented? A tool that fails these tests won’t improve once it’s in your stack.
Do you need a data strategy before an AI strategy?
Yes. There is no AI strategy without a good data strategy. AI can’t compensate for poor data quality, broken processes, or misaligned teams, and using it on a weak foundation makes existing problems worse. Fix the foundation first, then automate on top of it.
How should a small or solo team adopt AI?
Build the process before you add the tool. A solo marketer or small team should map the workflow they want to run, then use AI to execute it, rather than expecting AI to create the process. Onboard AI models the way you’d onboard a new hire, with clear guidelines and enough context to produce useful work.
Does AI replace project management and process?
No. AI runs a process well only when the process already works. It’s an accelerator layered on top of a sound foundation of clear workflows, good data, and defined accountability, not a substitute for building that foundation.
Workzone helps mid-market teams run their work on a foundation that’s actually built for how they operate. If you’re rethinking your processes before you layer AI on top, see how it works.
Last updated on July 8, 2026
Want a Peak Inside Workzone?
Continue reading
Ready To See Workzone In Action?
Simple project management software with real people ready to help you get started.
