Forty-six legal questions answered. Average response time: two minutes. User satisfaction: 4.2 out of 5. Two questions needed human follow-up - so she set a target to get that even lower.
A year ago, this would have seemed like a gimmick. Now it's just a Tuesday.
We build AI software. So, we had to prove it on ourselves first.
Ultimo makes AI-embedded Enterprise Asset Management (EAM) software for industrial businesses - manufacturers, logistics companies, asset-heavy operations trying to get more out of their equipment and their people. We'd been telling customers that AI would transform their workflows. That their teams would be augmented, not replaced. That it was time to move.
At some point, you have to be willing to go first.
Last year, our CEO set a clear mandate: full AI transformation across every department by end of 2025. Not a pilot programme. Not a proof of concept sitting in a sandbox. Real deployment, real accountability, real results. We hired an AI Transformation Lead, committed proper budget, and gave the project executive time.
What happened next surprised us.
The thing that changed everything wasn't the technology.
Early on, we built AI assistants that could handle questions and automate tasks. They worked. People used them occasionally. Nobody was particularly excited, and the impact was thin.
The problem wasn't what we'd built. The problem was that nobody owned it.
When the tool gave a bad answer, who fixed it? When it could do more, who pushed for that? The honest answer, most of the time, was nobody. And tools without owners don't improve, they stagnate.
So, we changed the model entirely. We stopped treating AI as technology we deploy and started treating it as people we hire.
Hunter, our account planning agent, has a profile photo. A job title. A place on the org chart. He reports to our Sales Director, who reviews his performance monthly, takes feedback from the reps who work with him, and makes calls about whether he should take on more complex work or needs his knowledge base updated.
Same with Harry in HR. IT-Cathy in IT support. MAX in marketing. Contract IQ in legal. Each one has a manager. Each manager is accountable for results.
We noticed the language shift almost immediately. People stopped saying "I used the AI tool" and started saying "I asked Harry" or "Hunter put that account plan together." That sounds like a small thing. It isn't. That shift is what turns occasional usage into daily habit.
Month one numbers, unvarnished.
Hunter created 333 account plans. Before Hunter, each one took a sales rep ten to 12 hours - researching the prospect, mapping their industry, tailoring the pitch, building the materials. Hunter does it in minutes. That's 3,500 hours back to the sales team in a single month.
Harry answered 229 employee questions about policies, benefits, expenses, and time-off procedures. Our three-person HR team used to field these constantly; now they have 20 hours back for the work they actually want to be doing. IT-Cathy handled 47 support queries, saving 10 hours. MAX built eight complete account-based marketing campaigns, saving 160 hours of copywriting and planning. Contract IQ's 10 hours brought the total to 3,720 hours saved across the organisation in month one.
But the number that stays with me isn't any of those. It's watching our sales reps stop working evenings. It's seeing HR finally redesign the onboarding experience they'd been meaning to get to for two years. It's legal having the headspace to focus on strategic partnership negotiations instead of fielding repetitive queries.
The hours are real. What people do with them is the actual transformation.
Three mistakes worth skipping.
First: don't hire an AI engineer as your first move. Hire someone who understands business processes - someone who will spend two months talking to department heads, mapping workflows, and finding where knowledge gets stuck and where the same questions get asked on repeat. The technology is the easy part. Knowing what to build is harder.
Second: don't skip governance because it feels slow. We spent three months building an AI Ethics Board, embedding AI Champions in each team, and running proper training before we scaled deployment. It felt like forever. But when data privacy questions came up, we had clear guidelines. When teams hit roadblocks, we had people who could help. The organisations struggling with AI right now are mostly the ones who skipped this step.
Third: don't make AI optional. We tried positioning our agents as helpful tools people could use if they wanted. Adoption was mediocre. The moment we flipped the default - Hunter does the first draft of every account plan, Harry is the first stop for HR questions - everything changed. The question shifted from "should we use AI for this?" to "why wouldn't we?"
What we're still figuring out.
We're five months into this. Twenty of seventy planned agents are live. The hard work - refining performance, expanding capabilities, managing change as workflows evolve - is mostly ahead of us.
Next quarter we're tracking employee NPS on AI usefulness, targeting above eight. That'll tell us whether our people genuinely value working alongside these agents or are just tolerating them. That distinction matters.
What's already clear is that the organisations that move now will have a meaningful advantage over those still deliberating. Not because the technology is magic - it isn't - but because the learning curve is real, the organisational muscle takes time to build, and there's no shortcut through it.
We sell AI-embedded software. We owed it to our customers to prove it works on ourselves first.
Now we can.