How PMP Strategy hit 97% AI adoption with Dust
- Industry
- Consulting
- Company Size
- 51-200
- Department
- Consultants
Key Highlights
- 97% AI adoption across the firm’s consultants
- 500+ agents created within 6 months by every consultant, building individual awareness and literacy
- A curated library of 50 high-value, firm-wide published agents
- Top management onboarded individually prior to firm-wide rollout
About PMP Strategy
PMP Strategy is a boutique strategy consulting firm of 150 people working directly with CEOs and senior executives on high-stakes strategic decisions. What sets them apart is the caliber of their advisors, how they support their clients through implementation and an internal Agentic Studio: a team of data scientists, data and AI engineers embedded. That team has been central to everything that followed.
The firm operates mainly across four sectors: Telecom, Media and Tech; Transportation and Mobility; Energy and Industrial Decarbonization; and Financial Services.
The challenge: the in-house platform hit a wall
PMPS was early to AI. In November 2022, when GPT-3.5 appeared, Gilles Vaqué, PMP Strategy’s Managing Partner, saw the potential immediately of how AI would disrupt every market and job. Data privacy and security being a sensitive topic in consulting, PMP’s Agentic Studio built an internal platform called “ChatPMP.” For a firm of 150 consultants, it felt like a competitive edge at the time.
For a while, it worked. Consultants used it to draft emails, translation, and basic document summaries. But the platform couldn’t keep up with the rapid evolution of AI.
“We were delivering one agent every four or five weeks, and even then, they were very limited. At some point, consultants came to us and said, this is not good enough.” - Olivier Leroy, head of AI
The problem was structural. Every new agent had to go through Agentic Studio. There was no way for consultants to build for themselves. The team had become a bottleneck, not an accelerator. As the pace of AI development accelerated, the gap between what was possible and what PMPS could deliver widened.
The firm needed a platform that would let everyone build.
Why Dust
- PMPS had a clear stance on AI platforms from the start: never become fully dependent on one vendor.
- The Agentic Studio wanted to maintain their development capability, using Dust as the primary entry point for consultants while retaining the ability to build more complex, proprietary agents as needed. Dust’s architecture made that possible.
- The harder test to pass was legal. PMPS works exclusively with client documents, which means confidentiality is non-negotiable. “We spent a lot of hours, both on our side and with the Dust team, reworking the contract to make sure everything was watertight,” Olivier says. “That’s a very important subject for us.”
Beyond the legal and technical fit, Dust being a French company, under EU AI Act regulation, was important.
The pilot was planned for 15 to 20 users and quickly expanded to 50. In PMPS's open workspaces, Dust went viral within weeks.
The solution: building a culture, not just a platform
Once the legal work was done and Dust was live, the real challenge began. Deploying a platform is straightforward. Getting 150 consultants, who have little time to spare to change how they work, is not.
“From the very beginning, two senior directors were involved,” explains Pierre (SD). “We created a working group of consultants and the Agentic Studio team, with a mix of seniority ranging from consultants to directors.”
Gilles, Pierre, Olivier and the AI committee members built a 3-layer adoption strategy from the ground up:
Starting at the top
The first move was to go straight to the partners.
“We held two-hour sessions, one-on-one, with each partner,” Olivier explains. “We took their notes, scattered post-its and reminders, and built a personal assistant to prioritize their workload. Then they say: “this is powerful. And the news spreads.”
When the most senior people in the firm adopt Dust, it send a strong signal to the rest of the organization.
"AI has immense transformative power for companies, which will need to rethink their business and rewire their processes. Executives must fully grasp this broad shift in the business model to coordinate efforts in the right direction. At PMP Strategy, half of our weekly Partner Meeting is dedicated to AI" - Gilles Vaqué, PMP Strategy’s Managing Partner
Building a network of champions
The second layer was identifying who in the firm was already leaning in.
Five former testers became natural AI champions after the test period. The Agentic Studio then tracked engagement through Dust's usage data and spotted the early adopters who built more than others after the rollout, and they became additional AI champions. The “Champion’s League” now meets bi-weekly.
"The internal AI champions became the ripple effect. Their close proximity to their colleagues, had more influence than top-down could ever have." - Olivier Leroy, Head of AI
Each Champion became a SPOC (single point of contact) for the high-value published agents. Their job is to listen to feedback, understand how agents are being used, and decide whether they need to evolve.
A governance structure that keeps delivering
The third layer is the AI Committee which meets weekly bringing together the CEO, a handful of partners, the Agentic Studio team, legal, HR, and one or two champions. The agenda never changes. Two items, every week.
First: which new agents should be published? Second: how many agents are being used across active projects?
"Gilles, our CEO, notices if we go a few weeks without new agents," Olivier says. "He asks: what's happening? We're not covering every possible case yet. We need to keep delivering."
Published agents carry a specific prefix, "PMPS_", so that any consultant, regardless of their familiarity with AI, can look at the name and description and immediately understand whether it applies to their work. Newly published agents are announced in a weekly internal newsletter. When the practice started, there were 11 published agents. Three deliveries later, the number was at 17, with a target of 50.
Embedding AI into every mission
Beyond the firm-level adoption, the committee tracks a specific metric: how many agents are used within each client project.
When a new project launches, a Champion spends an hour with the Project Manager to map out the use cases, which published agents apply, and which project-specific agents might be worth building. By week three of a typical six-week project, there is already a picture of which agents were created and whether any of them are strong enough to be promoted to the published library.
Pierre classifies agents into three categories:
- standard agents that come natively from Dust,
- PMPS’s published agents available to all consultants,
- project-specific agents.
“When someone builds a project-specific agent that has real impact, we bring it back to the committee,” he says. “We refine it properly, to serve a broad range of situations and contexts.”
It’s a virtuous spiral: consultants build, the best ideas surface through the committee, and Agentic Studio polishes them for firm-wide adoption.
Results: 97% adoption, 500 agents, and an entire firm that builds
97% adoption across PMPS’s consultants reflects how deeply AI is woven into how the firm operates.
- Over 500 agents have been created across the firm. The vast majority are exploratory, built by consultants learning what the technology can do. For Gilles, it’s intentional.
“We told everyone: build whatever you want. We may toss ninety percent of it. But you’ll have reached a level where you understand the tools, and when you have a real need, you’ll know how to build for it.”
- What Dust has changed for PMPS is not how much work gets done in a given hour, but where the work happens. Consultants now run complex exploratory analyses that would previously have been scoped out of a project. They can go deeper into document review, cross-check large volumes of materials automatically for consistency, and process information at a scale that was not possible before. For example, PMPS Deliverable Review, one of the published agents, flags gaps and inconsistencies in a presentation before it goes to the client.
"Dust gives us the time and perspective to step back and help our clients think through their strategic challenges," says Pierre. "That is where our value is, and where we want to spend as much time as possible."
- PMPS now uses its own adoption playbook with clients. Top management onboarding, champion networks, weekly governance, project-level tracking, is what PMPS recommends and implements for organizations going through their own AI transformation.
"The transformation is thirty percent technology and seventy percent governance," say Olivier and Pierre. "Technology is the easy part. What we built with Dust is a system that makes the hard part manageable."
Interested in learning more about how Dust can help your team? Visit our solutions page or reach out to our sales team.
