Most case studies about AI transformation come from a vendor describing its own product. This one is different. It’s an outside look at what happened when a small marketing agency decided to run its AI adoption the way a change professional would, with a defined case for change, a target state, and a benefits realisation case, before building anything.
Disclosure: Change Strategists was founded by Mark Draper, who is also the founder of Clixoni, the agency profiled in this piece. We’re naming that upfront because the connection is relevant, and because a case study is only useful if you can trust how it was reported.
We spoke with Draper about what the process actually looked like from the inside.
A familiar current state
By Draper’s own account, Clixoni spent most of last year looking exactly like the kind of business it was built to fix for clients. “A website on one platform, a content plan that existed mostly as an intention, a CRM that didn’t talk to the website,” he says. “I was the thing holding the seams together.”
Run a proper current-state assessment on that setup and it wouldn’t come back as broken. It would come back as what most businesses look like after a few years of evolving without a plan: functional, but inefficient, bolted together one decision at a time.
Draper says he sees the identical pattern constantly in client work, not just in his own agency. He points to a telecommunications provider he’s worked with as a specific example: a website on one platform, leads tracked in a spreadsheet by hand, no visibility into where a lead had actually come from or what they’d looked at before calling. “Different business, different sector,” he says. “Same current state.”
It’s a pattern change professionals will recognise immediately: not a crisis, just chronic inefficiency, the kind that rarely generates its own urgency.
The catalyst
What actually forced the question, according to Draper, wasn’t an internal failure but an external data point. A widely cited Forbes analysis published in July 2026 found that 95% of AI investments are currently showing no measurable return, with only 3% of leaders reporting they feel prepared to lead an AI-enabled team (Forbes, July 2026).
“Read that as an outsider and it’s just a statistic,” Draper says. “Read it while you’re running a business built on exactly the kind of bolted-together stack it’s describing, and it’s a mirror.”
Why now, and not two years earlier
AI tooling capable of writing a passable blog post has existed for a while. What Draper says actually changed wasn’t the ambition, it was whether the technology could be trusted with the daily grind rather than an occasional impressive demo.
There’s a meaningful gap between AI that can produce one good post on request and AI that can write, schedule, and publish in a client’s own voice every day without drifting into generic AI copy. The first is a demonstration. The second is infrastructure.
Notably, Clixoni had tried a more limited version of this before and pulled back. Social media posting used to be part of the agency’s offer, then was dropped entirely, because, in Draper’s words, “a promise that couldn’t be consistently kept by a small human team wasn’t a promise worth making.” It’s a useful marker for anyone evaluating AI change: a technology being capable in principle and being reliable enough to build a business process around are two different thresholds.
Crossing the first one too early is, by several accounts, how organisations end up in the position WRITER’s 2026 enterprise survey describes, with three-quarters of leaders admitting their AI strategy is more performative than functional (WRITER, 2026).
Designing the target state
The target state Draper landed on was simple to describe and, by his account, considerably harder to build: one system, with AI running the operational work end to end, and a human checkpoint before any of it reached a client.
In practice, that meant rebuilding rather than augmenting. Every piece of Clixoni’s daily work, website, blog content, social posting, Google Business Profile management, backlink outreach, and visitor engagement tracking, now runs on AI as the default execution layer. That’s the part most people picture when they hear “AI-first agency.” The less visible decision, and the one Draper returns to repeatedly, is the layer built around it.
The governance mechanism
That checkpoint, Draper argues, is the part any change professional should look for first, and the part most AI rollouts skip. “It’s not a nice-to-have bolted onto the plan for optics,” he says. “It’s the governance control that determines whether the rest of the system is trustworthy at all.”
At Clixoni, nothing goes out to a client without a named person checking it first, not a documented policy that quietly lapses under deadline pressure, but an actual gate, applied every time. Placeholder text, invented statistics, or anything that doesn’t sound like the client is meant to be caught there before it ships.
The stakes behind that design choice show up clearly in the wider data. Microsoft’s 2026 Work Trend Index frames the underlying problem as structural: only 19% of AI users sit in what the report calls the “Frontier,” where personal AI capability and organisational readiness reinforce each other, and organisational factors, not individual effort, account for roughly two-thirds of whether AI delivers real value (Microsoft Work Trend Index, 2026).
Automate the work without a governance layer holding it accountable, and the pattern WRITER’s survey found tends to follow: 29% of employees, 44% of Gen Z specifically, admitting to actively working against their own company’s AI rollout, not through simple disengagement but through shadow tools, misused data, and deliberately sandbagged output, largely because employees can tell when a strategy is more performative than functional and respond accordingly.
Draper is candid about one specific wrinkle in his own version of this. “Sponsor and delivery lead were the same person on this project,” he says. “Nobody signs off my own change programme but me. That’s its own kind of accountability, and it’s probably exactly why the checkpoint got built properly instead of left as a slide in a deck nobody revisits.”
The benefits case
Draper shared the kind of breakdown he says would normally go in front of a steering group before anyone signs off the spend. Sourced separately, the components that make up what Clixoni now calls its Growth Engine typically run:
- Website and blog: $25-60/month in tools plus the owner’s own time, or $2,000-5,000 for a one-off agency build.
- Blog content: $50-80/month in tools if written in-house, or $1,800-4,500/month for a content agency.
- Social posts: $50-120/month in tools, or $500-1,500/month for a social media agency.
- Backlink outreach: $150-300/month run in-house, or $500-2,500/month for a link building agency.
- Engagement tracking and CRM: $280-600/month across tools that mostly don’t talk to each other, or $700-1,500/month for a consultant to run it.
Add it up, and the current state costs somewhere between $555 and $1,160 a month managed in-house, or $3,500 to $10,000 a month paying agencies to run the pieces separately. Clixoni’s target state packages all of it into a single, managed system.
Scaling: from pilot to a second service line
Change programmes that hold up tend to get piloted before they get scaled, and Draper says this one followed that pattern deliberately. The Growth Engine came first: build the target state, run it reliably for a real client, and confirm the governance model held under actual use before offering it more broadly.
With that foundation proven, Clixoni’s second phase followed as an extension rather than a separate initiative: a lead generation layer, an AI assistant designed to turn quiet website visits into booked enquiries, built on top of the same governed system once a client already has it running. It applies the same underlying principle one step further along the funnel, catching visitors who show real interest but never fill in a form or pick up the phone, and starting that conversation before they leave the site.
What this means for other change professionals
None of what Draper describes is unique to marketing or to agencies. For anyone building a case for AI change inside their own organisation, three points from Clixoni’s experience seem worth carrying over regardless of industry.
The case for change rarely announces itself as urgent. Inefficient isn’t the same as broken, and most AI change gets delayed because nothing is technically on fire. Waiting for a crisis is usually just waiting.
The threshold that matters isn’t whether AI can do something impressively once. It’s whether it can be trusted to do it reliably, every day, without supervision quietly turning into hope. Those are different bars, and confusing them is how a promising pilot turns into next year’s version of the WRITER survey.
And the governance layer isn’t the part to cut when a timeline tightens. It’s the part that determines whether anyone downstream, employees, clients, or customers, actually trusts what the system produces. Skip it to move faster, and the more likely outcome is becoming the next data point in somebody else’s failure statistic, rather than the exception to it.
Readers curious to see the system described here in practice can find out more about Clixoni’s Growth Engine and lead generation services at changestrategists.com/website-services.
Enjoying our content? Make us a preferred source on Google.
Add us as a Preferred Source on Google
