The 20% Target That Became 80%
A Fortune 500 software company had a problem most enterprises will recognise. Leadership approved AI tools for the whole organisation. Budget was set. Licences were live. IT gave the green light.
Then almost nobody used them.
The company set a recovery target: get 20% of its 15,000 employees actively using AI within a few months — a modest, achievable number given where things stood. Within that same window, usage reached roughly 80%. Course completion topped 85%, well above anything the company had seen from a learning initiative before. Daily AI use climbed by 40 percentage points.
None of that came from a bigger training budget. It came from changing the starting point of the rollout entirely.
What was actually broken
The barriers weren’t technical. Roughly 40% of the workforce said they simply didn’t have time to figure out a new tool on top of their existing workload. Trust was thin — early, clumsy runs with AI had left people wary of the output. And without guidance tied to their actual job, most employees had no clear answer to the question: what does this do for me, specifically?
Every tool was already available. The gap wasn’t access. It was whether anyone believed the effort was worth it.
Why the Standard Training Playbook Couldn’t Have Gotten Them There
The instinct when adoption stalls is almost always the same: build more training. More modules, more sessions, more reminders to finish the course. It’s a reasonable instinct — and it’s also the reason most AI rollouts plateau well short of their goal.
Training is built to transfer knowledge. It assumes the person sitting through it already wants to use the tool, and just needs to know how. That assumption breaks down the moment someone is unconvinced the tool is worth their attention in the first place — which describes most employees at the start of any AI rollout.
Three things training was never built to fix
| What’s actually happening | What training assumes instead |
| Employees don’t have spare time to experiment with a new workflow | Employees have room in their schedule to sit through a module |
| Employees can’t picture how AI changes their specific job | A general feature walkthrough will make the connection for them |
| Employees have been burned by an unreliable AI output before | Explaining how the tool works will resolve any hesitation |
Two-thirds of employees who haven’t tried workplace AI tools point to time as the reason. Fewer than four in ten of those who do use AI say they’ve had any formal training at all — and many of that minority still can’t connect the dots to their own role. And the trust gap is wide: most desk workers say they aren’t fully comfortable relying on what AI gives them, which isn’t surprising given how few organisations have built any governance around when an output should be double-checked.
None of that is a knowledge gap. It’s a confidence gap — and confidence isn’t built by a tutorial.
| Time, relevance, and trust are feelings, not facts. A module can hand someone a fact. It can’t hand someone a feeling. That’s the gap a creative campaign is built to close. |
What the Campaign Actually Did, Phase by Phase
Instead of opening with a training calendar, the company opened with a story. Two phases, run in sequence, each solving a different half of the problem.
Phase one: make the tool feel worth trying
The first move had nothing to do with teaching anyone how AI worked. Five short, live-action videos were produced, each built around a character employees would immediately recognise — a developer drowning in code review, a marketer swamped with asset requests, an HR lead buried in candidate screening.
Each video followed the same arc: here’s the problem this person has right now, here’s the moment AI changes it, here’s what their day looks like afterward. The videos ran everywhere employees already spent their time — Slack, email, internal channels, team meetings — rather than living inside a learning platform nobody would think to open.
Nothing in this phase asked anyone to do anything yet. The only goal was curiosity.
Phase two: make the relevance impossible to miss
Once interest existed, a Digital Academy launched with five separate learning paths — one per major role group, each built around that group’s actual day-to-day work rather than a generic overview of AI capability.
A developer’s path covered code review and documentation. A marketer’s covered briefs and asset production. HR’s covered recruitment screening and internal communications. None of it asked employees to translate a generic lesson into their own context — that translation had already been done for them.
Sessions were kept short by design, directly answering the time complaint that had stalled adoption from the start. The Academy expanded over months with baseline content for everyone, deeper material for employees ready to go further, and a structured 20-week rollout to keep momentum from stalling.
| 80% active employees using AI daily, up from a 20% target | 85%+ course completion — above any prior learning benchmark | +40pts increase in daily AI usage across every business function |
| A training-only version of this rollout would likely have landed around 70–80% completion — and usage would have stayed flat. Completion isn’t the same as adoption. The story phase is what turned a finished course into a daily habit. |
| Stalled on Your Own AI Rollout? House of Designers builds the same story-first adoption campaigns for businesses across Orange County. Let’s talk about where yours is stuck. |
Five Companies, One Pattern: How Adoption Sticks at Scale
The Fortune 500 case isn’t an outlier. The same pattern — solve a real, felt problem first, let the tool prove itself there — shows up across very different industries.
| 01 | Microsoft — Copilot folded into the tools people already had open Rather than launching a separate AI product, Microsoft built Copilot directly into Word, Outlook, and Teams — the apps employees were already using all day. Search and writing tasks got noticeably faster, catching up on a missed meeting dropped from roughly 43 minutes to about 11, and the majority of users said they wouldn’t want to give the tool up. The lesson: adoption is easier when AI shows up inside an existing habit instead of asking for a new one. |
| 02 | Walmart — One expensive, unglamorous logistics problem Walmart didn’t pitch AI as a transformation story to its logistics teams. It pointed an in-house system at a single costly inefficiency — how trucks were routed and loaded — and let the savings speak. The result was roughly $75 million saved in a year and a meaningful cut in fuel-related emissions, with no separate change-management campaign required. |
| 03 | JPMorgan Chase — Automating the task lawyers already hated JPMorgan’s contract-review system, known internally as COIN, took on a job nobody enjoyed doing manually: combing through loan agreements line by line. Because the system replaced friction rather than adding a new tool to learn, the people it affected had no reason to resist it. It now handles work equivalent to roughly 360,000 hours a year, freeing legal teams for judgment calls AI can’t make. |
| 04 | CarMax — A volume problem no team could solve manually anyway CarMax had well over 100,000 customer reviews and no realistic way to process them by hand — manual effort would have taken over a decade. Generative AI summarised and organised the content in a matter of months, with no internal resistance to overcome because there was no manual alternative being displaced. |
| 05 | Shell — Shifting engineers from reactive to proactive, not replacing them Shell’s predictive maintenance system monitors more than 10,000 assets globally, flagging likely failures before they happen. Engineers didn’t lose their jobs to the system — they shifted from reacting to breakdowns to acting on predictions, which is a more interesting version of the same role, not a different one. |
What This Means for Your Own AI Rollout
Strip away the industry differences and five things hold true across every example above.
- The problem came first, not the technology: Every example started with something expensive, slow, or disliked — not with ‘we should use AI because it’s available.’ The tool was the answer to a question employees already had.
- Governance came before scale: Each organisation had a clear sense of when to trust an output and when to check it. Without that, trust never gets past the early hesitation stage.
- Humans kept the decisions that mattered: JPMorgan’s lawyers, Shell’s engineers — none of them got replaced. AI took the repetitive part; people kept the part that required judgment. That distinction is what makes adoption feel safe instead of threatening.
- Success was a number, not a feeling: Dollars saved, hours eliminated, percentage points gained — every case had a measurable outcome, which is what separates a real rollout from an expensive pilot nobody can defend later.
- The pilot was never the finish line: Shell didn’t stop at a handful of assets. CarMax didn’t stop at internal summaries. What worked small got pushed further, because the people closest to it actually wanted more of it.
| None of this requires a 15,000-person workforce to apply. The same five patterns hold at 50 employees as much as 50,000 — only the scale of the campaign changes, not the logic behind it. |
How House of Designers Builds AI Adoption Campaigns
House of Designers runs the same playbook described above for clients building their own AI rollout — strategy and creative production together, not handed off between two vendors.
01 Diagnose the actual barrier
Before producing anything, we find out whether your stalled adoption is a time problem, a relevance problem, or a trust problem. Each one needs a different campaign — guessing wrong wastes the budget.
02 Produce role-recognisable video content
Short videos built around personas your employees will see themselves in — not a generic feature demo. The goal is curiosity, not instruction, in this phase.
03 Build learning paths by role, not by topic
Separate, shorter paths for each function — developers, marketers, ops, HR — so nobody has to do the work of figuring out how a generic lesson applies to their job.
04 Launch where people already are
Slack, email, team meetings, internal comms — the campaign goes where attention already exists instead of waiting for employees to visit a training portal.
05 Track usage, not completion
We measure daily active use and time saved, not how many people clicked through a module. That’s the number that tells you whether the rollout actually worked.
Frequently Asked Questions
What is an AI adoption campaign?
An AI adoption campaign is a creative, story-led initiative designed to shift how employees feel about a new AI tool before they’re asked to learn it. It typically combines short-form video built around recognisable job roles with role-specific learning content, and is distinct from standard training in that it leads with persuasion rather than instruction.
Why did training alone fail to drive AI adoption at this company?
Training assumes the learner already wants to use the tool and just needs the steps. At this company, the real blockers were a lack of spare time, no clear connection between AI and an employee’s specific role, and low trust in AI output after earlier mistakes. None of those are knowledge gaps a module can close — they needed to be addressed before any instruction began.
How much did AI adoption actually increase?
The original target was a 20% increase in active AI usage across 15,000 employees. The campaign reached roughly 80% active usage, with course completion above 85% and a 40 percentage point jump in daily use — four times the original goal.
What made the campaign different from typical AI training?
It was sequenced in two phases instead of one. The first phase used short videos to build curiosity and emotional buy-in, with no instructional content at all. Only once interest existed did the second phase introduce role-specific learning paths, so employees encountered the ‘how’ after they already wanted the ‘why.’
Do the same adoption principles apply outside large enterprises?
Yes. The barriers — limited time, unclear relevance, low trust in outputs — exist in a 50-person company exactly as they do in a 15,000-person one. What changes at smaller scale is the size of the campaign, not the underlying approach.
How does House of Designers structure an AI adoption project?
We start by diagnosing which barrier is actually stalling adoption, then produce persona-based video content to build interest, followed by role-specific learning paths delivered through channels employees already use. We measure daily active usage rather than training completion, since that’s the number that reflects real behaviour change.
