Generative AI
Splio: Unlocking a projected 429% ROI with agentic AI
Companies have gotten very good at generating customer insights. Journey analytics, propensity models, sentiment tracking: on paper, most enterprises know their customers better than ever. Yet little of that knowledge is translating into revenue.
Research from the IBM Institute for Business Value found that only 34% of the customer data organizations collect today finds its way into CX decisions, while the rest remains locked away in silos. It also found that companies lose an average of USD 29 million a year in operating waste when they take too long to act on customer signals.
When that action gap persists, insights go stale, while the AI initiatives built to produce them get scaled back or quietly abandoned before they prove their worth. The result is missed revenue and opportunities captured by competitors, leaving companies with little to show for the insights they worked so hard to generate.
We'll explore the main causes of this disconnect, and how we helped Splio, a European customer intelligence leader, move fast to close it: in just three days, we built a working concept and a business case projecting 429% ROI.
The cost of doing nothing
When a company fails to act on its customer insights, nothing dramatic happens at first. No system crashes. No line item turns red overnight. The model that captures the insights just keeps running quietly in the background, still ingesting data, still consuming compute, still costing money, while nobody downstream does anything differently with those insights. We call this "the quiet loss."
This is the biggest problem I see: an initiative doesn't show results in the first month or two, so it gets abandoned. But the model itself doesn't disappear. It keeps running and consuming resources.
That loss compounds every month a company keeps paying for infrastructure that's technically working but practically idle, the AI equivalent of leaving the lights on in an empty office.
But that's not the whole problem. The bigger loss never shows up as a cost at all; it shows up as revenue that was supposed to happen and simply doesn't. For every insight a company generates but never acts on, the real bill isn't the price of the model. It's the price of everything that model was meant to protect or unlock, quietly not happening.
Why insight stalls before it becomes action
Typically, the problem isn't the model generating the insights itself. In almost every organization we've worked with, the intent, engineering and investment are there. What's missing is the ability to turn those insights into something the business can actually use.
Across our client work, we see three common gaps:
The data foundation isn't ready. Good insights need more than accurate data. They need the right context, connections and structure for a person or agent to understand them and act on them.
Nobody designed the workflow. An insight needs a clear path into the business: what happens when it appears, which decisions it informs and what actions follow. Teams often jump straight to building an AI agent without designing that process first.
Ownership stops at the handoff. The data team may own the model and the business team may own the channel, but neither necessarily owns the outcome in between. Someone needs to be accountable for taking the insight through the workflow and making sure it produces a business result.
Put together, these problems help explain why companies can end up with sophisticated insight engines and little to show for them in revenue. Closing the gap means addressing all three at once, which is the challenge we set out to solve with Splio.
Splio: Closing the gap in three days
Splio is a European customer-marketing company evolving from CRM and marketing automation toward a customer intelligence platform. A key part of that evolution is its focus on “individuation” marketing: using predictive customer insights to understand what each individual customer is likely to need or do next.
Splio already had the capability to generate those insights. The challenge was closing the gap between insight and action, giving marketers a way to turn those signals into targeted campaigns through a simple, conversational experience.
Through the AWS Agentic Catalyst Program and the 3/3/3 methodology, Thoughtworks, AWS and Splio tackled that challenge together. In just three days, we developed a working concept and a costed business case for taking it forward.
How the work came together
We started by mapping Splio's marketing value chain end to end. This helped us identify where insights were getting stuck and where an agentic AI solution could add the most value. We also surfaced the data issues that could prevent those insights from being used and outlined how a data-as-a-product approach could address them.
With that foundation in place, we moved into delivery. On Amazon Bedrock, we built a conversational, multi-agent AI workbench that combines predictive insights with a marketer's plain-language intent to create end-to-end campaigns, giving those insights a path to action.
By the end of the three days, Splio had a working concept demonstrating that the workflow was technically feasible, along with a business case projecting 429% ROI over three years and a four-month payback.
The speed was deliberate. So was the way we designed the solution. Three choices helped turn a promising concept into something the business could realistically take forward:
A workbench of specialized agents: Rather than one general-purpose assistant, we gave different agents a defined job. This makes the workflow easier to control, test and improve and makes it clear where to look when something goes wrong.
Brand safety by construction: We built compliance and brand-safety checks into the workflow, with human approval at each step. Brand guidelines and hard business rules are handled deterministically wherever possible.
- The 3/3/3 delivery motion: The three-day Catalyst work established the concept and business case. From there, Splio can continue the journey following Thoughtworks’ 3/3/3 method, with three weeks to test the prototype with real users, followed by three months to launch a minimum lovable product.
From insight to advantage
In three days, we had a prototype we could present to customers. Three weeks later, we were testing it internally and with early adopters. We are now industrializing it into a scalable product capability. We are extremely proud of how quickly we have moved from ambition to evidence, and from evidence to execution. The 3/3/3 approach gave us the structure to move at this speed without losing sight of trust, quality or customer value.
The takeaway
Companies can have the data, the models and the investment, yet still struggle to turn insight into action. The answer isn't necessarily better intelligence. It’s designing what happens next: the decision, the workflow and the ownership needed to turn an insight into a business outcome.
For organizations facing the same gap, a few lessons apply:
Start with the outcome. Define what business value the insight should create and how you’ll measure it.
Design the path from insight to action. Define the decisions, workflows and actions that need to happen before you build the technology.
Make ownership explicit. Give someone end-to-end accountability for turning the insight into a business outcome.
Prove value quickly. The goal isn't to spend less time thinking; it's to get to evidence quickly enough to make a better decision about what to do next.
Understanding your customers and what they need is the starting point. But insight alone isn't enough. What matters is what you’re able to do with that understanding.
