The Pattern Behind Every AI Disappointment
![[OPT 2] IMG CMO – Editorial – pattern behind every ai disappointment_-59514](https://thecmo.com/wp-content/uploads/sites/8/2026/07/OPT-2-IMG-CMO-–-Editorial-–-pattern-behind-every-ai-disappointment_-59514-792x446.jpg)
Momentum: AI offers quick wins but can exacerbate existing workflow issues without a solid foundation.
Workflow: Effective marketing requires prioritizing workflows over tools to achieve sustainable results.
Cost of Failure: Misimplementation and inefficiencies can lead to significant hidden costs beyond the tool's price.
Handoffs: Inefficient handoffs between teams can slow campaigns, diminishing the benefits of AI enhancements.
Redesign First: Before adopting AI tools, identify and address workflow breakdowns to ensure greater ROI.
The first thing a new marketing leader needs to show is capability. To build momentum, AI is an obvious lever, so building AI into workflows is a no-brainer. It’s fast to deploy, visible to the executive team, and easy to frame as a value-add.
Except, there's a sneaky ROI gap. And even with the tools in place, the results tend to lag behind the pitch.
The challenge for many marketers is that workflows are fragmented to begin with. Piecemeal platforms designed to serve a cohesive understanding fail to do so, and AI only speeds up the inaccuracies. Dysfunction isn’t any one person’s fault, all AI has done is put the organizational discourse under the microscope.
As a marketing leader, you may find yourself spending more time navigating individual use cases for AI, than being able to make the case for its ROI. And, you wouldn’t be alone.
Want more from The CMO?
Sign up for a free membership to complete reading this article:
Have an account? Log In
The pattern of selecting the tool first, and deciding on the specific friction points, runs across org sizes, platforms, and industries. The cost is hardly ever correctly attributed.
The Cost of Blaming the Wrong Thing
Every marketing team you know has an agent live somewhere in the workflow now. It's easy to blame the workflow first when all anyone is doing is prompting a chatbot time and again. People blame the platform or the vendor, then start evaluating alternatives.
Akande Davis has watched that exact sequence play out enough times at companies between $100M and $500M in revenue to know where it breaks down. A platform gets blamed for failures that started somewhere else, so the team migrates, and the same problems show up in the new tool a few months later.
"Nine times out of ten," he says, "it was implemented poorly, or the team wasn't enabled the way they should have been when purchasing the product."
Davis is VP of Operations at GNW Consulting, an agency that works with mid-market and enterprise B2B teams across Adobe, HubSpot, and Salesforce implementations.
His firm built a report quantifying what that cycle costs, and the sticker price of the tool turned out to be the smallest number on the page.
"The expense isn't just what you're paying for the product. It's the time lost, the revenue lost, the process inefficiencies—which all lead to a really large aggregated cost of failure that will unseat CMOs."
Why the Workflow Is the Hard Part
That mismatch shows up in fresh survey data too.
The State of Marketing report from Salesforce, surveyed 4,450 marketing leaders through November 2025 and found that only 13% of marketing teams are currently using agentic AI.
Even though most of those who use it, or plan to, expect real ROI gains from it. The gap comes down to workflow readiness.
High-performing marketers are 2.4 times more likely than everyone else to have unified their data sources. Plus, the average marketing org still needs to connect roughly seven different data sources before an agent can act on them reliably. An agent executing a broken process doesn't slow down to notice, it exacerbates the problem.
Liza Adams, a fractional CMO who advises companies through AI transformations, has a theory for why the workflow rarely gets touched even when everyone agrees it should.
"AI is not the hard part," she says. "It's the humans."
She describes a canyon sitting between two states, where a chatbox can answer anything, and a workflow where an agent is already built into how the work gets done end to end.
Training a team to prompt a model well is easy enough. But asking them to rebuild a process they've run the same way for years, and to trust that the agent running it still needs them, is a different kind of work entirely.
This is the layer most leaders underestimate. AI training tends to stop at features: what a tool can summarize, what it can draft.
Workers learn to use AI without ever seeing what their own role looks like once AI is built into the workflow instead of bolted onto the end of it. Adams calls that gap the actual roadblock, not the general resistance-to-change people usually blame.
Adams sums up what's changed between last year and this one in a single word: reimagining. She describes 2025 as the fast-and-furious year, when the mandate was just do something with AI.
The harder work starts now, when the question shifts from whether a team uses AI to whether that team can picture its own job differently once an agent is handling part of it.
"AI is now infused in how we do work," she says. "We can't do work without it." That's a slower kind of change than swapping tools, and it's the one most rollouts skip.
What Crossing That Gap Looks Like
Yael Abbukkis had already crossed that gap by the time she talked about it. The CMO at Lusha rebuilt her team over nine months around this same insight, restructuring from function-based tracks into workflow-based ones.
Her instruction to the team were simple, stating, "I don't want you to work on a brief. Instead, I want you to build a machine that makes marketing happen at 10x the speed."
The outcome: 16% ICP growth year-over-year while cutting paid media spend by 50%.
The result automated monitoring of competitor sites for unannounced UI and pricing changes. Plus, they released an end-to-end campaign system that turns a single input into deployable copy, creatives, and platform-formatted ad assets. ICP growth hit 16% year over year while paid media spend dropped by half.
Abbukkis knows what blocks other teams from getting there.
"Most leaders make the mistake of layering AI on top of a broken foundation," she says. "AI reflects the quality of the system it's plugged into. When data is fragmented and targeting is broad, AI just helps you move faster in the wrong direction."
Where to Start
You need to map where the work is breaking, not where you assume it is.
Define who owns each step before you automate anything further. Evaluate who requests the work, who reviews it, who signs off, and what stays human-led.
Once ownership and roles are settled, tool selection turns into a matching exercise instead of a bet. Abbukkis's advice for getting unstuck is actionable.
"Pick one workflow that matters and rebuild it end to end. Lead routing, ICP targeting, data enrichment, choose one. When one system works cleanly, the conversation about going further funds itself."
Abbukkis still calls her team's system a machine. She built it herself, mapped every handoff, before she let AI anywhere near it. Most of the teams running an agent right now could build the same machine, if they mapped the workflow before they turned it loose on one.
Those numbers didn't come from a better tool on it's own. They leveraged the tool to design a bespoke workflow system that reduced friction.
If you're wondering whether that kind of result is replicable, Abbukkis is direct about the single biggest thing that blocks it.
"Most leaders make the mistake of layering AI on top of a broken foundation. AI reflects the quality of the system it's plugged into. When data is fragmented and targeting is broad, AI just helps you move faster in the wrong direction."
Your competitors may be able to license the same tools but they can't replicate the workflows only you can build.
What’s Next?
Sign up for a free account at The CMO Club and get our newsletter delivered weekly straight to your inbox.
