Phil Turns One: A Year of Progress, Trust and Possibility

Our AI-powered economist, Phil, turns one this month, and this milestone is as much about Phil’s fanbase as it is about IBISWorld.

When we launched Phil last year, our vision was simple: take IBISWorld’s analyst-driven research and make it instantly accessible in the moments when professionals need it most. One year later, Phil has become more than a tool. He’s a trusted partner to consultants winning new clients, bankers evaluating risk, accountants benchmarking costs and analysts preparing for critical meetings.

Phil’s journey has shown us how powerful it is to pair human-driven research with AI speed, and how transformative it can be when that power is placed in the hands of professionals who need to act with confidence.

The road to launch: Listening to our clients

Before Phil ever answered a single question, we asked hundreds of clients how IBISWorld could better serve them. Their answers were strikingly consistent: our depth of research is unmatched, but navigating that depth could be overwhelming.

When large language models (LLMs) became viable, we saw a path forward. But we knew our approach had to be different. Clients rely on IBISWorld for objectivity and credibility, and we were determined to preserve that. That’s why Phil was built to draw only from IBISWorld’s proprietary database: the data and analyses crafted by our in-house experts. These guardrails and others, combined with rigorous testing, have helped us deliver the highest quality answers possible.

Phil is not your average chatbot. Where other LLM tools deliver unverified answers, or link to outdated sources, Phil is restricted only to the well-maintained, up-to-date insights that IBISWorld has become known for.

Behind the scenes, Phil uses natural language interpretation and advanced retrieval methods to surface the most relevant insights. For users, the effect is simple: answers that are fast, intuitive, and grounded in research you can trust.

Adoption at scale

The results speak for themselves. In Phil’s first month, he answered just under 2,000 questions. By July of this year, he was handling more than 50,000 in a single month. Adoption has grown ten-fold since launch, with 1 in 10 of our members now acquainted with Phil. This is proof that Phil is no longer a novelty but an everyday part of client workflows.

This growth reflects more than numbers. It reflects trust. Professionals are turning to Phil not just for quick answers, but for structured analyses, competitive insights and presentation materials they can take straight to the boardroom.

Real-world wins: How professionals use Phil

Phil’s impact becomes clear when we look at how clients are actually using him:

• Cutting research time: Consultants commonly prompt Phil for industry summaries that drop straight into pitch decks and proposals.
• Sharpening analysis: Analysts ask Phil to build four-part briefings (definition, demand drivers, performance, outlook) that mirror a mini-report in minutes, tailored to meet their exact needs.
• Clarifying the unclear: Even vague prompts like “buyer power” or “companies” return useful, targeted insights for a wide range of applications.
• Adding depth to decisions: Accountants and bankers lean on Phil to highlight competitive pressures, consolidation trends and cost dynamics that shape their strategies.

Even unexpected requests, from niche questions about regulations to multilingual prompts and replies, have shown how adaptable Phil is. He doesn’t just answer questions; he connects the dots in ways that make complex decisions clearer.

I remember how thrilled the team was to discover that Phil could speak Korean. We’ve long been aware that our reports get translated using Google’s translation features, but we also know that direct translations can sometimes overlook contextual clues and lead to misinterpretations. Being able to provide another tool for the many global users that rely on IBISWorld’s research is a huge relief. While AI translations have their limitations, Phil’s help is definitely welcome as we continue our own localized translation projects.

Learning and refining

Phil’s progress hasn’t been about perfection, it’s been about relentless refinement. Some experiments, like adding memory, didn’t improve accuracy. So we rolled them back and doubled down on precision.

After troubleshooting, we were able to reintroduce memory to specifically assist with follow-up questions. The result was an astounding 3x questions asked. We were thrilled to see that Phil could hold a conversation with our clients, allowing them to revise their prompts and expand their line of questioning in a natural and, most importantly, helpful way.

We also improved what happens when Phil can’t answer. Today, Phil operates with over a 93% success rate, which is up from the 80% range where we started. 

When we first released Phil to the world, our number one goal was quality. To make sure every answer was accurate in our research, we ringfenced Phil to a single report at a time. This safeguard gave clients confidence that they could trust his responses while we continued refining his capabilities behind the scenes.

But we learned that there were limitations to keeping Phil locked into one report and one report only.

With these strict rules, Phil was averaging a higher-than-expected fail rate. His failed responses often had to do with information just beyond his line of sight: the data and analysis trapped in our related Risk Ratings or ESG products, for example.

We were able to teach Phil a bit more about what IBISWorld offers in our full product catalog so that (even when hyper-focused on the IBISWorld data in front of him) he can still make recommendations and provide next steps for sourcing information outside of his purview.

Now, a banker or auditor looking for risk-based information, for example, knows that we have the answer somewhere else in our platform. And Phil can show them the way.

We’re proud to offer these flexible AI experiences, thinking always about the wide range of professionals we serve.

Flexibility was paramount when deciding exactly who should get access to Phil. With some organizations still cautious about AI, we knew that launching Phil would not equate to a flip of the switch for all of our clients. Instead, we knew we had to offer two tracks: AI-driven support when you want it, or a fully AI-free experience when that’s the requirement.

What’s next: From assistant to partner

Phil’s first year proved what’s possible. His second will be about deepening the value he provides.

Soon enough, those risk averse professionals I mentioned above will be able to query Phil directly on our risk-based insights. As we continuously expand Phil’s universe of knowledge, leaning always on our analysts’ high-quality research, we’ll be able to answer even more of the questions that busy professionals send his way.

We’re also rethinking how he accesses our information to begin with, considering a more sophisticated API approach that will yield speedier and more detailed responses. And we’re exploring how Phil can embed more deeply across the platform, so he’s present wherever you work, whether that’s in a Risk Ratings Report, our Segment Benchmarking product or in Enterprise Profiles.

As businesses continue to do more and more with AI, we see Phil becoming not just an assistant, but a partner. One that adapts to role-specific needs, supports self-education, and amplifies decision-making across workflows.