.png)
Servco's Peter Dooher on avoiding the AI quick-fix trap: why data foundation comes first, HR must lead transformation, and Frankenstacks usually fail.
For most marketing leaders heading into 2026, the pressure to show progress on AI adoption is real.
Most organizations respond by deploying AI agents across functions, presenting pilot lists in stakeholder meetings, and claiming integration. The result, according to Peter Dooher, Senior Vice President of Digital at Servco Pacific, is predictable: fragmented systems, pilots without success criteria, and teams moving on before learning anything.
Peter has spent two decades leading digital transformation at Accenture, J.Crew, and now Servco, Hawaii's largest automotive distributor and Australia's largest Toyota retail group. At Servco, he's retrofitting a 106-year-old company to operate AI-native, dealing with the exact challenges most enterprise leaders face: legacy systems, organizational inertia, and the tension between showing fast progress and building foundations that actually scale.
As one of Adora's earliest clients, Servco's transformation work has shaped how we think about what enterprise AI requires. In a recent conversation about enterprise AI transformation with Adora CEO Marco Matos, Peter shared where most AI initiatives go wrong, why treating AI as a bolt-on solution fragments your architecture, and why the most successful transformations are led by HR, not IT.
Here's how to drive AI transformation that actually scales impact, whether you're a century-old heritage brand or a digital-first challenger.
First, let's cover what Peter isn't investing in around AI.
"I'd have to say it's the quick fix. Stand up an agent and say, 'We check the box, we're running AI across the function.' Or the tendency to show a list of 1,000 things we're doing all at once."
The problem isn't that organizations are moving too fast. It's that they're treating AI as a quick fix for deeper operational problems, buying solutions without thinking through how those solutions fit into broader objectives or workflows.
What you end up with is a collection of AI-powered point solutions that don't talk to each other, built on different data foundations, solving tactical problems without addressing systemic ones. The Frankenstack (more on that later) emerges not from recklessness but from reasonable decisions made without a unifying strategy.
"It's not another add-on," Peter explained. "This is very different. It's not another bolt-on. This is ground up. This is rethinking the whole flow of your operations, your business, how you engage with your customers."
When organizations treat AI as something you can buy to fix specific problems without changing underlying workflows or data foundation, they skip the foundational work that determines whether those investments actually deliver value.
Peter is clear about the trade-off: "You may not see that outcome today, as opposed to a quick fix. But when it comes, it's going to scale, and we'll be able to measure it."
Organizations under pressure to show AI progress face a choice: buy quick fixes that fragment your architecture, or invest in integrated foundations that enable transformation that actually scales.
So if not quick fixes, where should a legacy organization begin?
"I think it's really important to start with the problem space," Peter explained. "Having studied physics and really focused on those first principle aspects of applying the scientific method, getting to the root of things, focusing on really important problems."
This first-principles thinking forces a different conversation than most vendor pitches. "Great vendors will talk to you about the features, and they've now got AI. But what are the real outcomes they're driving from those investments with their customer base, and how do we relate to that with our approach?"
Peter's filtering process begins with identifying genuine bottlenecks: "Where are the friction points in the journey today for the consumer? Really finding those friction points out, using the traditional value-ease method of prioritizing work."
Those specific, measurable problems, where manual processes slow execution, where teams reconcile data instead of driving growth, where customers experience friction, ensure your AI investment isn’t just a hammer looking for a nail…or a technology looking for application.
Those specific, measurable problems ensure AI investment makes sense and isn’t just a hammer looking for a nail…or a technology looking for application.
This is why MIT research found 95% of custom enterprise AI tools fail to reach production: the discipline to start with business problems, build measurement frameworks before investing in solutions, and focus resources on a manageable set of initiatives feels uncomfortable when everyone else is launching dozens of pilots. But it's what separates the 5% who succeed from the 95% who don't.
For Peter’s team at Servco, successful implementations are the ones solving clear business problems with measurement built in from day one. "We can show reduction in friction, we can show customer sentiment return. We know in the short-term it's probably redeploying some of those gains to different areas, but we know in the long-term it's going to show up in a more enjoyable experience for consumers."
Most vendors pitch AI as the solution to your reporting and insight problems. Consolidation through intelligence. Unified customer views powered by machine learning. Automated decision-making that cuts through data silos.
Peter's message is simpler and less comfortable: if you have bad data, AI tooling doesn't fix it. It just exposes and amplifies it.
"AI used inadvertently or not well could just amplify a problem in your data stack or martech stack," Peter explained.
The same data quality issues you've been working around for years suddenly become impossible to ignore when AI systems produce inconsistent outputs, conflicting recommendations, or results that clearly don't match business or brand reality.
Peter's approach starts with an unsexy question that most vendors skip: "Is [the data foundation] going to be one that we can rely on and trust and build quality off of? Then innovation such as AI can really amplify in the positive way."
An AI system built on fragmented data sources, inconsistent taxonomies, and unclear governance will fail regardless of how sophisticated the models are. The best AI in the world can't fix foundational problems; it just helps you discover them, but at the cost of time and investment.
For enterprise marketing organizations, this means the unglamorous work of data infrastructure comes before the exciting work of AI deployment. Customer identity resolution. Attribution methodology. Data governance. System integration. These aren't AI projects. They're prerequisites for AI projects that actually scale.
Before evaluating any AI vendor, audit your data foundation:
If not, that's where to invest first. When organizations skip this foundation work and jump straight to AI deployment, they inevitably create a monster of a problem.
This is where quick fixes compound into bigger problems.
“Fragmented AI is the one thing I want to be wary of. We can easily jump into [a scenario where] we have an agent for service and an agent for sales and an agent in the back office running on three different data sets. Fragmented parts almost Frankenstacked together."
The problem isn't deploying AI in multiple functions. It's deploying AI in multiple functions on incompatible foundations. Each system has its own data model, its own customer view, its own understanding of business logic. When you need to draw a cohesive customer journey across those systems, you discover the agents can't actually talk to each other in any meaningful way.
This fragmentation typically results from vendor acquisition strategies rather than intentional design. Companies buy point solutions that promise AI capabilities, bolt them onto existing stacks, and end up with systems that claim integration but operate in silos.
"Those who aren't rethinking it from the ground up, how their system stacks integrate, we find are really just trying to slap the label on top of it," Peter observed.
This Frankenstack dynamic isn’t new to marketing. But it is accelerating.
So how to avoid building Frankenstack? Look for partners thinking about end-to-end customer journeys, not just point solutions.
"We think of a value chain that goes well beyond the second or third purchase of that vehicle. All of that creates a really important data challenge to solve, and we have to make sure that the partners that we work closely with are thinking of it in the same strategic way."
That coordinated alignment matters more than any individual feature. Quick productivity gains are obviously valuable. But partners who succeed long-term are "building to the same unified goals that we are," understanding that today's tactical deployment needs to fit into tomorrow's integrated architecture.
But even with the right partners and unified architecture, there's one more consideration most organizations get wrong. And unfortunately, it’s the most important.
The most important insight: AI transformation is a people project, not an IT initiative.
"It's such a people-led change," Peter explained. "When we're going through our enablement, we're partnering head-to-head with HR. It's an HR-led initiative, because this is changing how we work."
This framing contradicts how most enterprises approach AI deployment. The initiative sits with the CTO or CIO. The technology team evaluates vendors, runs pilots, and handles implementation. HR gets involved later for training or headcount questions.
Peter flips that sequence. HR isn't downstream from technology decisions. They're co-leads from the start. "When we look at every role and function, it's going to be embedded in their workflow, and redefine their workflow, and actually redefine the operating model over time."
That scope of change can't be managed as a technology rollout. It requires organizational change disciplines: capability assessment, training design, change management, and culture transformation.
The challenge every legacy enterprise faces: "You were born an AI-first company, right? But we have to retrofit ourselves to become an AI-first company."
That retrofit touches every level. Servco stood up a Center of Excellence around AI, partnering with OpenAI to build capability systematically—board-level education, bi-weekly learning sessions across functions, connecting executive directions with ground-level implementation.
In organizations where AI falls to the technology team, rather than leading transformation themselves, you see executives asking for progress reports, "Where is our AI strategy? What are the results?"
Servco's approach makes organizational capability the priority, with technology choices following from clarity about what the business needs to do differently. This includes honest conversations about redeploying human capital. Legacy processes have people doing reconciliation across silos.
The goal isn't eliminating those roles. It's freeing that capacity for higher-value work. Moving from traditional brand marketing to growth marketing. Going after market segments the organization didn't have capacity to serve before.
As Peter described the Center of Excellence's role: "That's going to be what rewires us as an AI-first organization."
The rewiring isn't about installing new systems. It's about building new organizational capabilities, which is why HR needs to lead, not follow.
The path forward isn't complicated, but it requires saying no to most of what lands in your inbox.
Resist treating AI as a quick fix. Start with the business problems you're actually trying to solve, not the solutions vendors are selling. Invest in data foundation before deploying models. Choose partners building toward unified architecture, not vendors bolting AI onto acquired products. And recognize that organizational capability matters more than technology deployment. Meaning that if HR isn’t leading your AI adoption, you’re probably doing something wrong.
The 95% who fail chase quick wins that fragment their architecture. The 5% who succeed invest in foundations that enable transformation at scale.
Watch the full conversation between Peter and Marco to hear more on agentic AI, multi-modality, personalization at scale, and what Peter's specifically not investing in despite the hype.
