Esri's CIO Says Fix the Process Before Reaching for the Tool
- Kelsey Ziser
- 584 views
Esri CIO Thomas Phelps IV argues that process improvement, upskilling staff and keeping vendor knowledge in-house should come before adding new tools or AI to IT projects.

Thomas Phelps IV, CIO at geographic information systems company Esri, has a clear view on a habit many IT shops still fall into: picking a tool first and working out the problem afterwards. He calls that using technology as a sledgehammer. His alternative is to start with the business process, then decide whether technology, AI or neither is the right fix. Before joining Esri, Phelps was CIO and senior vice president of corporate strategy at Laserfiche, a director in the IT security, privacy and risk national practice at PwC, and part of the team that launched the Internet Software Products Division at Motorola.
Process first, then the tooling
Phelps says Esri has put real money and time into process improvement. The team recently hired and onboarded a process improvement specialist inside its IST organization. Phelps was on a kickoff call with her that same morning, focused on contract lifecycle management, and he came away impressed by how she ran the meeting and kept it pointed at outcomes.
His broader framing is people, process and technology, with culture and organisational structure added as a fourth and fifth element. Looking at the lifecycle of a business process alongside the specific technology area it touches is, in his view, the right approach. Esri applies that lens as it works with stakeholders in global sales, marketing, and finance and accounting.
The scale of the work matters here. Phelps says IT is managing about 38 discrete projects at once, many of which affect different parts of the business and involve customers. Executing those projects well isn't enough by itself, he argues. The team also has to look for ways to automate and to work differently.
Why automating a bad process backfires
Phelps puts it bluntly: you can always automate a bad process and simply do the bad process faster. His team's goal is to understand what each project is trying to accomplish, then build a solution that scales and addresses the root cause of the problem, rather than the symptom someone noticed first.
That's the practical reason the process lens matters. A workflow with unnecessary steps, or a gap where information gets lost between teams, won't be fixed by dropping in a new platform. It will just run the same friction through a different interface. Phelps's test is to ask how the process should be reengineered and reinvented, and only then how technology and AI can help it perform properly.
My read is that this is a sensible discipline for any team under pressure to show AI progress. The temptation to announce an AI pilot before anyone has mapped the workflow is strong, and the result is often an expensive automation of steps that should have been removed.
Don't outsource the knowledge
Asked about what hasn't gone to plan, Phelps pointed to a lesson he says most CIOs learn the hard way: over-reliance on systems integrators and third parties to deliver specific initiatives. He doesn't reject outside help. He wants clarity about what the third party is delivering and assurance that the work lines up with the right outcomes.
His approach is to assign a team, either newly stood up or reallocated, to work closely with the vendor during delivery. The aim is institutional knowledge of that product so the internal team can support the business after the engagement ends. He gave ERP upgrades and new security implementations as examples where this matters most.
Here's the thing: that kind of knowledge transfer is easy to promise in a statement of work and hard to deliver in practice. It needs named people with time set aside, not a shadow team that disappears when the contract closes. It also needs to be planned in at the start, not added as an afterthought.
Budget for the change you know is coming
Phelps says initiatives should be planned on the assumption that outcomes will shift. Staffing changes and scope changes are normal on any project. His advice is to budget for the extra time and cost at the planning stage, so a team doesn't discover halfway through that the money has run out.
He also recommends vetting the people who will deliver the work, not just the organisation. That means looking closely at the experience and skill sets of the individuals assigned to the project, not only the firm's credentials.
Where the talent money goes
On talent, Phelps described two parallel tracks with the employee development and training group. The first is annual leadership training for IST staff. The second is a technical skills review: finding out what the team already knows and then deciding how to build on it, such as courses on a particular vendor's AI stack.
He says the most important investment, though, is capacity. One of his goals this year is giving employees time to learn, experiment and apply what they learn to initiatives that deliver the right business outcomes. Many IT teams say they value learning, but few allocate the hours to back it up. Phelps is betting that protected time will pay off in better project results.
AI governance: agents need an inventory
Phelps says the rapid pace of new AI models requires his team to keep watching how they affect both Esri's products and its internal operations. The aim is to find practical uses for AI across the business in a way that is governed, safe and responsible, and his team is focused on using AI to run business operations at scale.
On the governance side, his concerns are specific. First, observability and a control plane for every agent running in the environment. Second, an inventory of those agents: what each one does, and what permissions and roles it holds. He calls that inventory critical to AI governance. Third, agentic systems need the right guardrails and controls enforced before they run.
For admins, that list translates into a fairly concrete audit. Do you know which agents exist, who owns them, and what they can access? If the answer is unclear, that's the first gap to close before the next rollout.
Reading list and a leadership framework
Phelps recommended "Scare to Cares" by Saby Warriach, a book about leading digital transformation without chaos. He pointed to the fear many employees feel about AI and job disruption, and asked whether leaders themselves are creating stress and anxiety that turns into resistance to change.
His answer is a CARE framework. Communicate with empathy. Adapt with agility. Build relationships based on trust. Empower the team with that trust, within the right governance, guardrails, coaching and mentorship. And stay calm through change. He also named Stephen Covey's "The 7 Habits of Highly Effective People" as his favourite book, citing its principles-based approach to leadership.
What to take into your next change window
Phelps's message is consistent. Map the process before choosing the tool. Keep knowledge inside the team when vendors are involved. Budget for change up front. Inventory your AI agents and their permissions before they multiply. And protect time for people to learn. None of it is new, but the AI wave is making teams skip these steps, and that's where the sledgehammer habit comes from.