Wicked Reports turns GenAI from prototype to revenue with DoiT’s Accelerator
Meet Wicked Reports
Wicked Reports, based in Waltham, Massachusetts, is a multi-touch marketing attribution platform that uses first-party data to give its customers more accurate information about their paid marketing activities in online advertising. It’s designed to optimize paid online advertising spend and effectiveness and primarily used by e-commerce brands with revenue ranges of $5 million to $50 million . Their goal: Reduce wasted ad spend.
“The number one reason customers choose us is accuracy,” says Wicked Reports CEO and co-founder Scott Desgrosseilliers. “If our data isn’t more accurate than what you get for free from the ad platforms or Google Analytics, there’d be no reason for us to exist.”
The Challenge
Wicked Reports knew using AI could shorten the time to value for customers navigating multi-touch attribution’s inherent complexity, but early experiments faltered. Early AI models hallucinated, failed to track time progression, or forgot prior recommendations, occasionally suggesting the opposite days later.
Guardrails were hard to enforce, and there was no secure, production-grade data pipeline. “We didn’t need AI for the sake of it,” Desgrosseilliers says, “we needed AI that delivered business value.”
Internally, the team’s mental model for AI had to evolve. Wicked Reports initially utilized AI to replace code-driven if-then logic, rather than synthesizing multiple signals into a coherent analysis with explanations written in clear, actionable language.
“We realized AI’s strength wasn’t flowchart logic, it was connecting twenty data points at once and explaining what to do,” Desgrosseilliers says. “The AI, if properly designed, it could follow our framework consistently in minutes.”
The Solution
Starting with the right program and the right incentives
The engagement began under DoiT’s Cloud Accelerators: GenAI program for AWS, which paired funding with a structured path from ideation to implementation.
“The accelerator was critical,” Desgrosseilliers says. “It allowed DoiT to put in the level of effort we needed.” That combination of AWS funding plus DoiT’s Cloud Accelerator’s clear framework turned a great business idea into an executable plan.
Translating the attribution framework into a production-ready agent
Wicked Reports provided comprehensive documentation that described a comprehensive decision framework, including the rules, thresholds, and caveats behind its mulit-touch attribution recommendation and optimization methodology.
DoiT’s technical lead, working closely with Desgrosseilliers and Andy Nadler, Product Manager at Wicked Reports, ingested the framework and iterated quickly: sample data was reviewed through prompts, followed by internal review, and then real client feedback on which recommendations were realistic in practice and which needed refinement.
“We were always moving the ball forward,” Desgrosseilliers says. “You can tell when someone really knows what they’re talking about. That expertise and the pace built our confidence.”
A major hurdle was prompt engineering for reliability. Wicked Reports needed the model to produce strongly typed, correctly formatted JSON every time so product engineering could consume results without hand-holding.
“We could have hit LLM endpoints ourselves,” Nadler says, “but consistently getting the same structure back every time was the part we couldn’t do. DoiT solved that.”
Building a secure AWS path from raw signals to guided actions
Under the accelerator, DoiT implemented a secure, serverless-first pipeline on AWS, API orchestration with Amazon API Gateway, event processing with AWS Lambda, containerized workloads on Amazon ECS with AWS Fargate, observability in Amazon CloudWatch, permissions via AWS IAM, secret management in AWS Secrets Manager, and infrastructure as code in AWS CDK.
The design respected Wicked Reports’ data sensitivities, connected to existing sources, and created a safe environment for rapid iteration.
“We didn’t just glue a model onto our product,” Desgrosseilliers says. “We created a real system.”
Turning “hallucinations” into governed intelligence
DoiT redesigned how the agent reasoned about time and context. The model began with a summarized history, then selectively drilled into high-signal periods for detail, avoiding the twin pitfalls of drowning in data or missing nuance.
Guardrails captured the team’s framework as an explicit policy, and the agent explained why a recommendation aligned with the method.
“It stopped being a black box,” Desgrosseilliers says. “It became our methodology at machine speed.”
Partnering like a product team, not a ticket queue
The working style mattered as much as the architecture. DoiT integrated with Wicked Report’s Slack, providing fast feedback loops and allowing for iterative changes – without waiting for weekly status meetings.
“They were highly responsive,” Nadler says. “We’d review results, give feedback, and see improvements right away.”
The Results
Three months of development time saved and opportunity cost avoided
By utilizing DoiT’s expertise to build the prototype and carry it to a production-ready state, Wicked Reports saved at least three months of development time, avoided pulling core engineers off the roadmap, and reduced opportunity costs.
“We would have lost a quarter and still not had the same quality,” Desgrosseilliers says, “Instead, our team kept shipping while the AI came to life.”
From concept to a sellable revenue opportunity with a clear go-to-market
The output of Wicked Reports’ engagement wasn’t just a lab demo; it’s a productized feature with real-world revenue growth potential. Wicked Reports launched the AI capability as a paid add-on with a time-boxed free trial, and the company is now ramping marketing and adoption.
The initiative has already transformed how the customer success team engages with Wicked Reports’ customers, with plans to have CSMs run the analysis and send results proactively for retention.
“We’re giving customers a taste that proves the value,” Nadler says. “The sales motion is built into the experience.”
Reliable, explainable recommendations that match the framework
The new agent consistently follows Wicked Report’s methodology, produces repeatable output, and explains how it arrived at its guidance. The change is tangible in day-to-day work.
“The AI does the right steps every time,” Desgrosseilliers says. “It’s our framework executed in minutes, without human drift.” The guardrails and JSON rigor also unblocked Wicked Reports’ engineering team, who can now integrate results directly without one-off formatting fixes or fragile manual checks.
A better way of working with a technology partner
Wicked Reports’ previous reseller experience left the team wary. The contrast with DoiT – expert, responsive, and accountable, became part of the result.
“They were a true partner,” Nadler says. “They helped us figure out which problem to solve and then built the solution that actually solved it.” Desgrosseilliers summarizes it simply: “Responsive, trustworthy, productive.”
What's Next?
The roadmap expands from tactical media guidance to broader strategic analysis. Wicked Reports plans to extend the agent to executive-level summaries, creative diagnostics, and anomaly tracking, then empower customer success to deliver that intelligence in ways that drive retention and upsell.
The company maintains an annual consulting relationship with DoiT and intends to keep iterating with the same people and cadence that made the first phase successful.
“We’re going to keep working with the team that shipped this,” Desgrosseilliers says. “They understand the problem space and they move.”