Digital Impact
Enterprise AI — Boomi CEO Steve Lucas, with CIOs and executives from Ford, Deloitte, Red Hat and the Red Cross, on why AI projects stall on disconnected systems and data, and what the teams that got it working did differently.
Mechanisms
Ask them for the car, not the chassis
The American Cancer Society's CIO, Ricky Koch, builds small, complete integrations with a little data first.
"“We’ve learned to bite off smaller pieces with our integrations, get them working well, and then expand." "When they ask for a car, they want to see a car. They don’t want to see a chassis and then hear you say ‘once that’s right, we’ll put on the hood, the engine, and everything else.’"
Why it matters
The American Cancer Society's CIO, Ricky Koch, builds small, complete integrations with a little data first. Mistakes surface early, and business teams see something that works. A half-built big system shows them nothing they can use.
Use when: You're building something large for non-technical stakeholders who need to see progress.
Let manufacturing see the design early
Fields found pockets of Ford each analyzing only their own data.
"When we started on digital transformation at Ford, I created a single data and analytics group." "So, if the engineers had this great new idea for an engine, transmission, or HVAC system, the manufacturing folks could say, “Hold on, if you design it that way, it will cost more, and servicing it will be harder.”"
Why it matters
Fields found pockets of Ford each analyzing only their own data. Pooling it let downstream teams flag costs before designs were locked. Catching the problem late had meant redone work, extra cost and slipped timelines.
Use when: Teams downstream discover problems only after the upstream team has committed.
Look for the damaged plant, not the bug
Nature Fresh Farms' camera robots couldn't find pests, which hid under leaves and fled the noise.
"“The moment this robot makes noise coming through, it just flies away. So, we decided not to look for bugs hurting the plant and look for plants impacted by bugs. That’s always a sure-tell sign where they are.”"
Why it matters
Nature Fresh Farms' camera robots couldn't find pests, which hid under leaves and fled the noise. Switching the target from the elusive cause to its visible effect made detection work.
Use when: You keep failing to detect a cause directly and its effects are easy to see.
Concepts
Data is the new sand
Oil is scarce and data isn't.
"In my opinion, “data is the new oil” was always a simplistic and flawed concept." "There’s no gold star for the business with the most data."
Why it matters
Oil is scarce and data isn't. Lucas argues most stored data is sand: disconnected, unused and a storage cost on the ledger. Value comes from precise, accurate slices put to work, which calls for planning around demand instead of hoarding supply.
Use when: Your organization is collecting everything in the hope it becomes useful later.
Evidence
Ship a limited scope, then iterate
The Credit Union of Colorado switched from multi-year launches to small releases.
"“Previously, we were built for large-scale project implementations. We’d take three to five years and roll them out only when they were as close to perfect as we could get them." "We’ll take a limited scope, implement, and iterate. We move a lot faster.”"
Why it matters
The Credit Union of Colorado switched from multi-year launches to small releases. It built integrations nearly five times faster, and its new digital member experience contributed to a 5% leap in NPS scores. Aiming for perfect at launch had mostly made launches late.
Use when: Your team holds releases until they're nearly perfect and projects take years.
Efficiency isn't always the right measure
Kalyra, an aged-care provider, nearly bought software to cut client phone calls.
"They want to pick up the phone because that might be their weekly conversation with somebody." "Efficiency isn’t always the right measure for success. Asking, ‘What is the value that we’re adding?’ is a better approach."
Why it matters
Kalyra, an aged-care provider, nearly bought software to cut client phone calls. A regional manager pointed out that many clients live alone. For them the call was part of the service, and cutting it would have saved staff time at the cost of what clients valued.
Use when: You're automating away a customer touchpoint and measuring only time saved.
Bad data is worse than none
Nature Fresh Farms' IT head, Keith Bradley, records every metric on its greenhouse plants.
"We’ve learned firsthand that you can have years and years of data, but if it’s not in a usable state and accurate, then it’s no good."
Why it matters
Nature Fresh Farms' IT head, Keith Bradley, records every metric on its greenhouse plants. He warns that inaccurate historical data is even more of a hindrance than none, because models trained on it go wrong with confidence.
Use when: You're about to train or analyze on a large historical dataset nobody has audited.
The tier-two supplier is what shuts you down
Ford bought $90 billion a year in parts from suppliers.
"In my time, I never had a tier-one supplier ever shut us down. It was always the tier-two and tier-three suppliers that shut us down because we didn’t have good visibility down there and never saw it coming."
Why it matters
Ford bought $90 billion a year in parts from suppliers. Risk hid where it couldn't see: its suppliers' suppliers. Fields would point a monitoring agent at their financial health and news.
Use when: You're assessing supply or vendor risk and only looking one level down.
Ninety percent of your apps nobody approved
The result was siloed data, poor visibility and security risk.
"One notable example of digital chaos was that 89% of the 200-plus applications at Australian Red Cross consisted of “shadow IT” – deployed without the technical team’s knowledge or oversight."
Why it matters
The result was siloed data, poor visibility and security risk. The overhaul started with finding out what was actually running, then replaced the tangle with seven core systems and one connectivity layer, and cut about 10% of its applications.
Use when: You're planning a technology overhaul and assuming your inventory of tools is complete.
Contradictions
My systems would work well if you stopped using them
Quinlan uses the joke to describe how IT departments used to measure themselves.
"They were set up to take care of technology, not people. People were an unnecessary distraction. The old IT joke was, “My systems would work well if you stopped using them.”"
Why it matters
Quinlan uses the joke to describe how IT departments used to measure themselves. Today almost everything a CIO wants to do depends on other business units, so a team built to protect its systems from users fails at its job.
Use when: A support function is optimizing for its own metrics at the expense of the people it serves.
Disruptive and overhyped at the same time
Vishal Sikka, SAP's former CTO, holds both views.
"It’s both incredibly disruptive and overhyped. It’s possible to believe both things simultaneously." "The hallucination problem will not just go away. It’s fundamental to the nature of the technology."
Why it matters
Vishal Sikka, SAP's former CTO, holds both views. AI can't do certain things people assume it can, and it is still enormously powerful. Holding both keeps you from overbuying and from dismissing it.
Use when: A debate about AI has split into hype versus skepticism and you need a working position.
Questions
Ask whose life gets easier
Larry Quinlan, Deloitte's former global CIO, says the implementations that succeeded were the ones tightly tied to the people using them.
"Instead, ask yourself, “How will this initiative make life easier for the actual people who have to use it?”" "It actually creates failure – expensive failure."
Why it matters
Larry Quinlan, Deloitte's former global CIO, says the implementations that succeeded were the ones tightly tied to the people using them. Digital for its own sake gets redone, delayed and abandoned, and careers stall with it.
Use when: You're approving a technology project justified by the system rather than the user.
Ask the CIO: what's the status of our data?
Former Ford CEO Mark Fields applies "garbage in, garbage out" to AI.
"It would be a very clarifying question for a CEO to ask their CIO: “What’s the status of our data?”"
Why it matters
Former Ford CEO Mark Fields applies "garbage in, garbage out" to AI. Many companies announce AI plans without knowing what data they have or where it lives. One question exposes that gap before money is spent.
Use when: Leadership is excited about an AI initiative and nobody has checked the inputs.
How many more beakers for five dollars?
Koch translates back-office savings into the mission's own units, research beakers and lodge nights for patients.
"“My whole goal is to reduce IT as a cost center and ensure more money is put into the mission,” he said. “I think about it in simple terms. How many more beakers can you buy with an additional $5? How many more nights can we provide in a lodge stay if I save $100?”"
Why it matters
Koch translates back-office savings into the mission's own units, research beakers and lodge nights for patients. It makes cost-cutting concrete for everyone and ties a support team's work to the goal.
Use when: You need a team to care about costs that seem far removed from the real work.
Passages
Moving out of mom and dad's house
A simple, shared metaphor turned a technical carve-out into a story everyone understood.
"“We described it as ‘moving out of mom and dad’s house.’ It pulled the team together and unified a clear intention in building our own capabilities – built by us, for us.”"
Why it matters
A simple, shared metaphor turned a technical carve-out into a story everyone understood. Lischett paired it with two fixed principles, simplicity and speed, and the team met its timeline.
Use when: You're leading a team through a complex separation or launch and need everyone pointed the same way.
Predictions
We always underestimate the rollout
Former Red Hat CEO Paul Cormier compares AI to Linux, when there weren't enough engineers for every company to hire its own.
"Three to five years. We always underestimate how long it will take to get these technologies into the real world and consume them."
Why it matters
Former Red Hat CEO Paul Cormier compares AI to Linux, when there weren't enough engineers for every company to hire its own. Skills, not the technology, set the pace. He still says to invest and experiment now.
Use when: You're setting expectations for when a new technology will pay off.
Rules
The three Ps, and the biggest P
Quinlan's test for a transformation: what will people get from it, how can you prove they're getting it, and did you ask whether they liked it.
"I would advise always thinking about the three “Ps” – people, process, and platforms. By far, the biggest “P” of all is people."
Why it matters
Quinlan's test for a transformation: what will people get from it, how can you prove they're getting it, and did you ask whether they liked it. Without those answers, he says, projects flounder.
Use when: You're planning a rollout and the plan is mostly about the platform.
Be willing to destroy the path
Bradley says the farm has failed many times on AI projects.
"The biggest stigma with AI is thinking that when you’re going down a path, you have to stick to it. We’ve learned that sometimes you have to destroy that path, go back to the data, and start rebuilding it in a new way."
Why it matters
Bradley says the farm has failed many times on AI projects. Treating each approach as disposable kept failures cheap, and the farm's yield runs eight to ten times traditional farming per square meter.
Use when: A project is failing and the team feels committed to the approach it started with.
Change management starts at go-live
Tropicana had about 18 months to build a full IT operation when PepsiCo sold it.
"Lastly, change management doesn’t end with the change itself. In some regards, that’s when the change management starts. Prepare for the first six months similarly to the previous six months before a go-live."
Why it matters
Tropicana had about 18 months to build a full IT operation when PepsiCo sold it. Its CIO, Jeff Lischett, budgets the same effort after launch as before it, because adoption problems only show up once people use the new system.
Use when: You're planning a big rollout and the plan ends on launch day.
Know what you're building before you pick up the hammer
Cornell's integration architect, John Parker, found the hard part was matching data to satisfy parties who each wanted it their own way.
"Everyone thinks, ‘My data is my data, and that’s exactly the way I want it.’ My advice is to get your arms around your data and decide what you want and can do before you get started."
Why it matters
Cornell's integration architect, John Parker, found the hard part was matching data to satisfy parties who each wanted it their own way. Agreeing on scope first avoids building something no side accepts.
Use when: You're starting a project that merges information owned by several teams.
Connections
Every system thinks it's the smartest in the room
Abhinav Gaur, who speaks four languages, manages integrations at Tony's Chocolonely.
"“Every system, probably just like every person, probably thinks that they are the smartest one in any conversation. But you need integration, sitting between them, smiling, and making their interactions work."
Why it matters
Abhinav Gaur, who speaks four languages, manages integrations at Tony's Chocolonely. He compares systems to people who each believe their own grammar is correct. The fix is a go-between that lets each side keep its own terms.
Use when: You're mediating between teams, partners or systems that each insist on their own way of working.