Join 250,000+
professionals today
Add Insights to your inbox - get the latest
professional news for free.
Join our 250K+ subscribers
Join our 250K+ subscribers
Subscribe06 JAN 2026 / TECHNOLOGY
Boston Consulting Group (BCG) has adopted a tech-forward approach to consulting, utilizing artificial intelligence (AI) in its operations. Through BCG X, a 3,000-person unit of engineers, designers, and data scientists, the consulting firm has built custom AI tools to assist in areas such as research synthesis and supply chain diagnostics; this tech-centric approach has resulted in productivity gains of 15 to 30% in client projects and saved employees significant time that is reinvested in higher-value tasks.
Picture a BCG partner at 11:47 p.m., coffee going cold, not tweaking a slide headline but stress-testing an AI agent. Not for fun. For a client deadline. Somewhere along the way, the classic consulting playbook picked up a compiler, a UX backlog, and a security review checklist. And now here we are. Boston Consulting Group, a firm old enough to remember overhead projectors, is acting less like a slide factory and more like a product shop. Not as a side hustle. As a core operating model. The message is blunt: if you are not tech-ready, you are not really ready.
BCG did not wake up one morning and decide to cosplay as a software company. The shift has been steady, intentional, and expensive. In late 2023, the firm formalized its AI push through BCG X, a 3,000-person build unit that blends consultants, engineers, designers, and data scientists. Around the same time, BCG rolled out ChatGPT Enterprise to its entire workforce. The result? More than 18,000 custom GPTs built internally in under a year. Some figures put the number north of 30,000, depending on how you count reusable agents versus one-off builds. Either way, that is not dabbling. That is industrial output. Why does this matter? Because it signals a deeper change. Advice alone no longer cuts it. Clients want tools that run on Monday, not just decks that sound smart on Friday.
BCG’s AI operation runs like a three-layer stack. At the bottom is data. Lots of it. Decades of presentations, project documents, transcripts, and research, all sanitized and structured into knowledge graphs and vectorized libraries. The focus is not just on smarter models, but on smarter data usage. Knowing what to pull, when to pull it, and why it matters. Accountants will appreciate this instinct. Garbage in still means garbage out. In the middle layer, things get scrappy. Consultants on live client engagements build AI agents to solve real problems. Pricing analysis. Research synthesis. HR self-service. Supply chain diagnostics. When a tool works, it does not die in a project folder. It gets sent back to the R&D team, reviewed, hardened, and added to an internal marketplace.
Roughly 80% of BCG’s custom GPTs come from this frontline work. Not from a central ivory tower. That bottom-up model is doing a lot of heavy lifting. At the top sit firm-wide tools. Think Deckster, an AI-powered slide builder trained on 800 to 900 approved templates. Or Ava, an internal assistant who handles IT and HR questions. Or GENE, a conversational AI used for podcasts, presentations, and thought leadership. These are not demos. They are production tools with dedicated teams behind them.
Here is the unglamorous part that actually matters. Every AI tool goes through red teaming, data privacy checks, legal review, and information security assessment. No shortcuts. No cowboy coding. Once approved, tools land in a central repository. An orchestration agent helps employees figure out which tool to use and which data source fits the job. BCG even runs an internal enablement network of about 1,000 people across HR, legal, finance, and operations. They train teams, collect feedback, and surface new ideas. Monthly. Like clockwork. This is what product governance looks like when services firms stop winging it.
Let’s talk impact, because vibes do not pay invoices.
Deckster alone has been used over 450,000 times since its global rollout in March 2024. Heavy users save two to three hours per deck. Formatting defects are down about 35% . Managers spend less time fixing fonts and more time thinking. That is a win any CFO can understand. GENE, built on GPT-4o, acts as a thinking partner for consultants. It ingests entire reports or podcast transcripts in one go, no external database required. Teams use it to brainstorm storylines, generate summaries, and co-host podcasts. Internal metrics show a threefold increase in citation depth compared to standard AI queries. Translation: answers got better, not just faster. Across the firm, BCG estimates that about 70% of the time saved through these tools is reinvested into higher-value client work. Not fewer hours. Better hours. That distinction matters.
This is not just internal optimization theater. Client projects using BCG-built AI agents report productivity gains in the 15 to 30% range. In one widely cited case, an OpenAI-powered customer service deployment reduced resolution time from 11 minutes to under two minutes and replaced work equivalent to 700 full-time employees. The projected profit upside was about $40 million in a single year. Those numbers tend to get boardroom attention. Fast. And then there is CO2 AI, the carbon management platform incubated inside BCG and spun out in 2023 with $12 million in funding. It now tracks over 400 million metric tons of CO₂e across more than 100 multinational clients. That is more than global aviation emits annually. The platform produces audit-ready Scope 1, 2, and 3 data, with full lineage. Accountants, meet your new sustainability workpapers.
Here is the uncomfortable question. If a 60-year-old consulting firm can rewire itself into a tech-forward product organization, what excuse does anyone else have? Every company now touches software. Even if you sell widgets. Even if you file returns. Even if you think tech is someone else’s department. The line between operations and engineering is blurry, and it is not snapping back. This does not mean every firm needs to build 18,000 AI agents. But it does mean treating technology as infrastructure, not garnish. It means governance, controls, audit trails, and human-in-the-loop design. It means asking harder questions about data quality and accountability. It means accepting that “we are not a tech company” is starting to sound like “we do not really plan to scale.” As the saying goes, you do not have to outrun the bear. You just have to outrun the slowest camper. BCG is clearly not trying to be the slow one.
Until next time…
Don’t forget to share this story on LinkedIn, X and Facebook
Subscribe now for $199 and get unlimited access to MYCPE ONE, from CPE credits to insights Magazine
📢MYCPE ONE Insights has a newsletter on LinkedIn as well! If you want the sharpest analysis of all accounting and finance news without the jargon, Insights is the place to be! Click Here to Join
You’ve reached the 3 free-content piece limit. Unlock unlimited access to all News & CPE resources.
Subscribe Today.
Already have an account?
Sign In