I like founders who aren’t distracted by the sparkle. Shelley Copsey is one of them.
She started in consulting on large infrastructure projects, moved from Australia to the UK, and, five years ago, took an investor-backed prototype and shaped it into FYLD — an AI-enabled platform now used by major utilities in the UK and North America. The team is in the mid-fifties and growing; momentum has shifted from early pilots to wider enterprise rollouts. She’s practical, people-centred and clear about outcomes: safer sites, fewer delays, better first-time fix.
As CEO and co-founder at FYLD, she’s building in an industry most tech people ignore — utilities, construction, infrastructure — and she’s doing it the hard way: field-first, with a product that fades into the background so crews can actually do the job. Talking to her provided me with some fantastic insights – here are some of the most exciting.
You can’t fix what you can’t see
It seems crazy, but we still run critical field operations with after-the-fact paperwork and WhatsApp archaeology. Then we act shocked when jobs stall, crews wait around, or risk bites us. Shelley’s lived this world for years. Her blunt assessment: the industry doesn’t have a “skills shortage” as much as a visibility problem. FYLD routinely sees standing time around the 30–35% mark. That’s not because crews are lazy. It’s because the system sends people out without the right kit, permits, or plan, then leaves managers blind until something goes wrong.
The fix isn’t another dashboard in head office. It’s line-of-sight to reality – what’s actually happening on site – and the ability to intervene in the moment. Not next week at the steering meeting. Now.
Shelley’s approcah is simple: put signal where the work happens. Give crews an app they’ll use in the rain at 2am. Let them capture short videos, photos, voice notes. Run AI over that stream so a command centre can spot risk and unblock delays. Surface exactly one or two things someone needs to act on. Then get out of the way.
That’s the model. Boring in the best possible way.
Safety versus productivity is a false choice
Leaders in construction still talk like it’s 1978: “We can be safe or we can be fast.” Shelley’s view: nonsense. Yes, sites today have more compliance and complexity. And yes, you “lose” 20–30 minutes at the start of a job making it safe. But the cost of incidents — stoppages, investigations, injured people — dwarfs that time. Safer sites go faster over the week, because flow beats firefighting.
Two practical points from Shelley:
- High-energy hazards are easy to miss. Crews aren’t scanning for suspended loads or stored energy the way a machine can. If tech highlights the one thing most likely to hurt someone right now, you reduce harm and delay with a single nudge.
- Cognitive overload kills action. Tell a crew they’ve got 24 hazards and they’ll process none of them. Give them the one that matters most and what to do about it — you get movement.
Outcome: when FYLD customers reduce incidents, capacity appears in the same system. Regulators love that. Finance loves that. People getting home safe at the end of the shift love that.
Adoption beats innovation (every time)
Founders love to talk “innovation”. Shelley’s allergic. Her order of play is adoption, then outcomes, then innovation. In field operations, the frontline decides who wins: if crews hate your tool, it dies quietly; if they like it, usage compounds and value snowballs.
That’s why FYLD was built field-first with hundreds of frontline workers (SGN among the early partners), so the app feels more like a simple banking feed than enterprise software — a clear timeline, read ticks, and quick video or voice capture instead of clunky forms. The technology is designed to be invisible: no one on site cares which model spotted the risk, only which trench needs attention and what to do next.
Even pricing removes friction — no per-seat rationing to throttle usage, just enterprise-wide access so capture can happen anywhere, any time. Do that and satisfaction shifts from grudging compliance to genuine, voluntary engagement — and FYLD’s high customer satisfaction from people in hi-vis who don’t normally hand out gold stars tells you the approach works.
AI: Avoid the doom loop, find the useful loop
We touched the “AI kills all jobs” narrative. Shelley’s take: the next decade will be bumpy and uneven — change arrives city by city, use case by use case — but the apocalypse talk is a distraction. Your job as a leader is to build the version of the future you want inside your firm, not wait for think-tanks to agree.
Start by picking one ugly, frequent workflow and make it smoother this quarter — method statements, risk assessments, job packs, first-time fix. Standardise language so the model learns from clean inputs: decide what “ready”, “standing time” and “high-energy hazard” mean, write it down, and use it everywhere. Close the loop on the ground so when the system spots something, the crew sees the nudge in the app they already use — no PDFs, no inboxes. Then measure outcomes, not theatre: fewer rework visits, more first-time fixes, less standing time, safer sites. If those don’t move, your AI hasn’t landed. Boring wins.
From pilots to platform: The enterprise shift
FYLD’s at an inflection point. Early adopters and experiments have given way to mass rollouts — including a US customer with tens of thousands of field workers. That shift forces discipline.
Shelley’s three moves you can copy:
- System over heroes. The “hire a hero and let them do what they like” phase doesn’t scale past five sellers. Shelley drove CRM discipline and robust qualification. If there’s no business case, no real champion, and no line of sight to value, it’s out.
- ICP clarity. Not every enterprise can ingest AI-enabled change. Disqualify faster. Walk away sooner. You’ll grow healthier and your reputation improves.
- References as currency. Infrastructure is conservative but collaborative on safety. Great outcomes travel by word of mouth. That beats any sales pitch.
Under the bonnet, FYLD is growing ~80% year-on-year and gearing up for the next funding step. Nice validation also comes from inclusion in lots of “hottest AI startup” lists — but what matters more is sticky adoption and expanding footprints. Revenue follows usage in this model.
Hiring leaders when ambiguity is the job
Sales leadership is hard in an early company. Ambiguous targets, long cycles, complex buyers. Shelley’s honest about the mis-hires and what she changed:
- Set the bar in the first sentence. She opens interviews with: “My default answer is no. You’ll need to prove you’re the best person in the world for this job.” It saves months. If that puts someone off, they weren’t your person.
- Culture beats credentials. At 50 people, one toxic “high performer” can wreck metabolism. Fast exits for culture breaches. No “brilliant jerks”.
- Hunt through your board. Her standout CRO came via board network, not a job ad. A good board doesn’t just opine; it opens doors and lends credibility.
One more leadership truth from Shelley: if you don’t trust someone in a pivotal seat, listen to your gut. Move them or move them out. You’re betting the company otherwise.
Board craft without theatre
Founders often accumulate advisors like novelty mugs. Shelley did the opposite. She picked a chair she clicked with on day one — industry-literate, calm under pressure, no ego — and added independent experience when the go-to-market needed sharpening. Investors have representation, but the board’s job is clear: make the CEO’s job easier and decisions better.
Tempo is a choice (and a habit)
Shelley runs hot. She protects the metabolic rate of FYLD and expects leaders to do the same. A few habits worth stealing:
- Ruthless prioritisation. Her unread email count is six figures. Not a boast, a boundary. Email is a bad way to steer a high-tempo company. Critical issues find her through the right channels; everything else can wait.
- Induction that actually indoctrinates. As you scale, half your team will be <12 months’ tenure. “Sheep dip” new starters in values, cadence, and the definition of done. Model it in public. Every day.
- Lead indicators, not lagging drama. Don’t wait for churn to tell you a customer was in trouble. Instrument your journey. If usage dips, exec sponsorship goes quiet, or value stories dry up — intervene this week, not next quarter.
- Read to raise the bar. She’s into Frank Slootman’s Amp It Up for pace and standards, and Oliver Burkeman’s 4,000 Weeks to avoid confusing busyness with impact. Good combo.
Tempo isn’t an accident. It’s daily choices.
What changes when you actually see?
When you install line-of-sight to real work, queues shrink because a missing part is flagged at 9:10 rather than discovered at 11:50. Risk starts to be managed rather than merely reported, with the meaningful conversation happening on site today instead of buried in next month’s pack. Capacity appears as regulators get more output from the system you already run and crews complete more in fewer visits.
Pride returns too: people like doing a good job and being seen doing it; hidden work demoralises, visible work motivates. You don’t need a moonshot to get there — just a clear, boring foundation of common definitions, effortless capture, managers who act on the signal, and pricing that encourages usage. Do that once, then again, then again.
Shelley described most people — especially the ones who keep our water running and gas flowing — as wanting to do a good day’s work and get home safe. Your job as a leader is to make that easier. Remove friction. Surface the right risk at the right time. Show them their work counts.
Wins that make a difference
Shelley’s approach is a reminder that the wins that move the dial rarely look like exciting. They look like an app that tells a supervisor, “Check that pipe.” They look like 40 minutes of standing time disappearing from a morning. They look like a parent finishing a shift intact because a nudge arrived at the right second.
If you run a business with any kind of field, plant, or frontline reality, start where she did: make the tech invisible and the work visible. One workflow, one crew, one week. Then repeat.