This article was originally published on LinkedIn.
By Jag Gattu, CEO & Co-founder at UptimeAI
A Board Built for the Benefit of our Customers
We write software for operations we don’t run. That makes staying grounded in how plants actually operate essential. The real test is not whether a product performs well in a demo or checks every procurement box. It is whether it works under real operating conditions, with real constraints, for the people making decisions every day.
So earlier this year we started an Industry Advisory Board — an independent group of senior operators who advise us personally based on their experience, not on behalf of their employers. We weren’t looking for endorsements. We wanted people who would tell us when we’re wrong.
How does this benefit our customers? When we set our priorities for the year, they’re aligned with the priorities of the major industrial players. When we decide to build a new agent or decline to build something, that decision has been judged by a room that includes people who have run refineries, generation fleets, and technology organizations.
Our Members
The initial membership includes:
- Matt Ballew, Director of Emerging Power Technology and Innovation at Vistra
- Darryl Willis, Corporate Vice President for Energy and Resources at Microsoft, with prior roles at BP and Google Cloud
- Mike Swanson, twenty-five years in downstream refining at ExxonMobil, CVR Energy and Big West, now consulting for America First Refining as they work to build the first new U.S. refinery in fifty years
- Jim Heilbron, Senior Vice President of Technical Shared Services at Southern Company
Meeting #1 – What We Learned
Matt, Darryl and Mike joined our first session in August. With no slides and a loose agenda, it left plenty of room to pull at the threads of every conversation. There was a healthy balance of alignment and disagreement. Here were my main takeaways:
- No one benchmarks their data readiness Darryl was candid that most of the data estates he encounters are in poor shape. Most companies lack formal processes to assess data readiness, which means it’s often overlooked when companies are assessing new technology solutions. Matt offered a practical rebuttal that aligns with a core belief we have at UptimeAI: if your people already make decisions with a set of data, and someone owns it, it is probably good enough to start with, particularly where a human stays in the loop.
- Half the market carries scars from AI pilots Asked how many of the customers he works with have been burned by an AI pilot, Darryl said half. He attributed it to three things — a dilapidated data estate, vendors who over-promised, and picking the wrong place to start. It means we’re starting from behind but acknowledging that early gives us the opportunity to correct perception early.
- A mandate is permission to change, not guaranteed change Mike made the point that the impetus must come from above. You shouldn’t expect an “underswell” of frontline supervisors advocating changes that challenge their status quo. Matt challenged that a mandate is only permission. What decides whether anything changes is a champion amongst those frontline supervisors, willing to iterate and not afraid to fail.
- Valuing technology is critical, and complicated One of the questions I raised was, “how do you put value on a failure that never occurred?” We didn’t come to a firm answer here. Matt called anomaly detection “fire insurance” and said it has always been a hard case to make: heat rate and megawatts tie to defensible numbers, while a prevented outage invites the question of whether it would have been caught anyway. Darryl reframed it around elapsed time and accelerated capital planning— his team cut permitting for a small modular reactor by 92% — and the team was able to align on the value of bringing the system online sooner.
I’m excited to dive deeper into this last topic in our next meeting. There are layers of value to a solution like UptimeAI and the experience of our IAB will help frame how we peel back the layers in each of our customer deployments.
