Inside BCG's AI product assembly line: 'Every company has to become a tech company'

• Boston Consulting Group is investing heavily in developing new AI-based tools.
• Consultants at BCG have built tens of thousands of custom AI agents for client projects.
• BCG relies on a bottom-up model, sourcing AI innovations from frontline consultants.

When it comes to creating AI tools, Boston Consulting Group is "all in," Scott Wilder, a Dallas-based partner at the firm, told Business Insider.

About 15 months ago, the firm launched an internal research and development lab — one of several ways in which it is developing new AI-based technology to facilitate its work.

"Every company has to become a tech company," he said. "BCG is no exception."

Ever since OpenAI unveiled ChatGPT in 2022, generative AI has become the shiny object companies can't ignore. And consulting firms from McKinsey to BCG — once known for their strategy and advisory services — are shifting more toward building, implementing, and maintaining AI tools for clients.

Wilder, who studied computer science at the University of Texas - Austin, said that "as a computer scientist, it's fascinating to see a services firm shift this way."

'Forward Deployed'

At BCG, generative AI innovation now happens at three levels, Wilder said.

First, there's the data level, which is driven by data engineers. "We're in a phase where we're really starting to leverage our internal proprietary data," Wilder said.

Wilder said BCG is building MCP servers and agents on top of its internal and selected public data sources so AI tools can automatically pull the right information at the right time — a shift he sees as the future of AI, where model intelligence is increasingly tied to knowing which data to use and when.

Then there is a "middle layer" of innovation that comes from consultants building tools and agents for client projects, Wilder said. When one of these tools is a hit, it goes back to the R&D team, which integrates them into a central marketplace for the rest of the firm to use.

Lastly, there's a top layer of executives developing firm-wide tools.

These include Deckster, a slideshow editor trained on 800 to 900 slide templates that helps consultants quickly create presentations, which also has its own development team. There's also Ava, an internal help tool for IT issues, HR questions, and other internal requests. There's also GENE, a conversational chatbot the firm uses for presentations, podcasts, and publicity initiatives.

Wilder said the firm has about seven or eight tools at the top level.

The firm also has a team of "forward-deployed consultants" — a role inspired by the forward-deployed engineers popularized by Palantir — who are vibe-coding and building tools for clients, Wilder said.

"We put our best of the best globally on cases, and they build reusable things. That then goes back to the R&D team," he said. "So we're seeing a lot of innovation happen from that. A lot of it has to happen at the rock face, on the case teams, versus centrally."

These consultants rely on a mix of external coding platforms and internal resources to build tools, including OpenAI's ChatGPT. BCG bills itself as the No. 1 creator of custom GPTs globally, and has created about 36,000 so far, Wilder said.

The new assembly line

Simultaneously, the firm's new R&D team works on disseminating these innovations internally.

When the R&D team launched about 15 months ago, its mandate was to build knowledge graphs from the firm's slide decks, transcripts, and other internal documents. Knowledge graphs are typically used to organize and connect data across multiple sources. So, with a meeting transcript, for example, the R&D team would aim to extract not just words, but also concepts and sentiments, Wilder said.

Now, the team works on fleshing out ideas that consultants working with clients on the front lines generate.

Wilder said that new tools are also screened by BCG's red teaming team, which simulates adversarial exercises to address a tool's security vulnerabilities. It is also reviewed by the data protection office to ensure compliance with data privacy laws, undergoes a legal review, and is assessed for information security.

Once it's ready, it's passed on to the firm's marketplace team, which integrates it into a central repository of tools for the rest of the firm to use, Wilder said. "There's actually an agent," he said. "This orchestration agent, basically, helps people understand what are the best tools and data within BCG to use."

Wilder said the firm prides itself on a "bottom-up" approach to innovation. "Probably 80% of our custom GPTs come from the front line," he said.

The new ideas also surface through the firm's "enablement network." This is a team of about a thousand people from every department, from human resources to legal, that functions like learning and development coordinators, training the rest of the firm on AI.

They also double as "our eyes and ears," Wilder said.

The team meets monthly to share ideas and insights they've gathered through their work with various teams.

BCG now operates much like a traditional product organization, Wilder said.

Between its dedicated product teams and a "UX Center of Excellence," which constantly evaluates what's working and identifies which agents still need to be built through frontline user research, BCG feeds those insights back to product owners, who then funnel the feedback into a centralized crowdsourcing pipeline.

"In short," Wilder said, "we're doing all the things you'd expect a product organization to do."