How Spendesk turned 90% adoption into embedded AI workflows with a champions program

Industries: B2B SaaS
Financial Services
Company Size: 201-1000
Departments: Company-Wide AI Adoption
Operations

Key Highlights

93-94% monthly active users across the entire company
83%+ weekly retention — most employees use Dust at least once per week
40%+ of messages sent to custom agents (not generic LLMs)
13 AI Champions driving adoption across every department

This story was written by Wesype, a Dust Platinum Partner with strong expertise in the FinTech industry. It shows how Spendesk moved beyond initial AI adoption success to build sustainable, process-embedded usage across all departments. With Wesype's support, Spendesk structured its Champions Program and achieved industry-leading retention metrics.

About Spendesk

What happens after you hit 90% AI adoption? For most companies, that's where the success story ends and where the real challenges begin.

Spendesk is a European spend management and procurement platform powered by AI, trusted by thousands of businesses across Europe. The company combines payment cards, expense management, invoice processing, procurement, accounting automation, and corporate travel management (launching soon), in a single platform. With offices in Paris, London, Berlin, and Barcelona and over 400 employees, Spendesk also owns its own licensed Payment Institution, making security and compliance non-negotiable.

By late 2025, Spendesk had already achieved what most companies dream of: 90% AI adoption within six months of deploying Dust. But the team responsible for sustaining that success knew the hard work was just beginning. High adoption metrics look great on dashboards, but the real question was: are people using AI in ways that actually transform how they work?

Challenge: When high adoption isn’t enough

The "Day 2" problem

Spendesk's initial Dust deployment in February 2025 was deliberately limited, around 20 users identified as potential future champions. By May, the platform had expanded company-wide, and by June, monthly active users had already hit 65-70%.

But there was a problem hiding inside that early success.

The initial limited rollout had created an unintended perception. With only select people having licenses, some employees assumed AI tools were only for certain roles or technical profiles when in reality, use cases existed for every department.

"We regretted not deploying to the whole company right away. Having only some people with licenses created the impression that only certain people could use it, when actually there were use cases for every department."
— Cécile Hervouet, Revenue Operations / AI Program Manager

The hackathon hangover

That summer, Spendesk ran a hackathon that generated real excitement. Around 50 participants created 11 new agents in a burst of creative energy.

Then reality set in.

Six months later, only one of those 11 agents was still being used regularly, and even that one had just a handful of messages per month. The momentum from the hackathon had evaporated almost immediately.

This pattern is common: companies celebrate AI "wins" that fade within weeks. The challenge isn't generating excitement, it's building sustainable habits.

From experimentation to infrastructure

The question facing Spendesk wasn't whether people would try AI. It was whether AI would become embedded in actual workflows, part of how work gets done, not an optional add-on that enthusiasts use when they remember.

Solution: a champions program built for sustained adoption

In September 2025, adoption was high, but depth was uneven. Some power users were building sophisticated agents; most employees were using Dust as a better ChatGPT.

The team's approach: stop chasing company-wide metrics and start embedding AI at the team level.

The Champions Program structure

Spendesk formalized what had been informal: a structured Champions Program with clear roles, dedicated time, and real accountability.

12 Champions across all departments:

Minimum one champion per department
Larger departments (Engineering, RevOps) have two or more
Champions officially allocate a minimum of 10% of their time to Dust initiatives

Three core responsibilities:

Evangelize: Help their team understand the ease and value of Dust
Build: Create agents for department-specific use cases and workflows
Report: Share best practices with other champions and surface insights to leadership

"I really feel that all champions have taken ownership of their role. They are well-identified within their teams as the champions. It took time to secure their bandwidth, but the sentiment is that they've all embraced it." — Cécile Hervouet

Governance that doesn't get in the way

Rather than heavy process, Spendesk implemented lightweight but consistent touchpoints:

Bi-weekly meetings (30 minutes): Champions share current use cases, ask questions, learn from each other
Monthly 1:1s with AI champions program lead: Deeper dives into each champion's department-specific challenges and opportunities
Quarterly SteerCos: Strategic alignment with leadership during flagship initiative phases

The governance was designed to support champions, not burden them—creating a rhythm of sharing and learning without bureaucratic overhead.

Learning from failure: The 6-week hackathon

After the summer hackathon's "one agent surviving" outcome, the team tried something different: a six-week "disguised hackathon" in November-December 2025.

Instead of a burst of energy over a few days, this format:

Re-assembled teams around agents created in the summer to see if they could be revived
Allowed time for iteration, testing, and feedback
Included space for new agents addressing emerging use cases

Results: 8 agents created, 3 still actively used weeks later. Not perfect, but a significant improvement in retention compared to the traditional hackathon model.

The insight: sustainable adoption requires sustained attention, not just initial enthusiasm.

Results: from adoption metrics to workflow integration

Custom agents became the default

By December 2025, the most telling metric wasn't monthly active users. It was that 40%+ of all messages were going to custom agents rather than generic LLM interactions. This number reveals something important: employees weren't just using Dust as a ChatGPT replacement. They were using purpose-built agents created by Spendeskers for Spendeskers, tools that solve specific problems in specific workflows.

Industry-leading retention to back it up

The depth of usage showed up in the retention numbers too. By December 2025, Spendesk had achieved metrics that most AI implementations never reach:

93-94% monthly active users across the entire company
80%+ weekly retention, with most employees using Dust at least once per week
13 Champions driving adoption across every department

"We hit 80% monthly adoption by November, then 93-94% in December. And for weekly retention, we're above 80%. That means almost everyone at Spendesk uses Dust at least once a week."
— Cécile Hervouet

Spotlight: The agents that stuck

AskProduct — Spendesk's most-used agent

The AskProduct agent allows anyone to query the company's product documentation and roadmap instantly. What makes it work:

Exceptionally well-documented knowledge base
High output quality with minimal hallucinations (only 2 reported since September)
Strong initial communication that drove word-of-mouth adoption

360 Customer View — Unifying siloed data

One of the first agents Cécile built connects data that traditionally lives in separate systems: Salesforce, product databases, and other sources that don't naturally talk to each other.

"For having worked a lot on customer data reconciliation, even in my previous roles, it was always a major challenge. Seeing that I could now do it myself, with no technical skills, was mind-blowing."
— Cécile Hervouet

Sales workflow agents — Organic adoption from top performers

The Head of Sales recently organized a workshop to build sales-specific agents—an initiative that emerged organically rather than being mandated from above. When top performers share their wins publicly (including on LinkedIn), others follow.

What makes agents survive

Not every agent succeeds. Based on Spendesk's experience, the agents that achieve sustained usage share common characteristics:

Excellent documentation: The knowledge base must be complete and well-maintained
Strong launch communication: People need to know the agent exists and what it does
Clear value proposition: Solves a specific, recurring problem
Champion ownership: Someone is accountable for maintaining and improving it

What’s next: from AI adoption to AI operations

Building a dedicated squad

Cécile is advocating for a formal "AI in Operations" squad, a small team with clear ownership of Spendesk's AI infrastructure for internal operations:

Core team: 2 people (one program manager, one technical specialist)
Extended support: Champions across departments + occasional consultant support for specialized projects
Potential addition: A part-time enablement/training specialist

The goal isn't to centralize all AI work, but to provide the infrastructure and support that allows department-level AI initiatives to thrive.

2026 roadmap: Process-level integration

Rather than chasing more agents, the 2026 focus is on deeper integration:

Process mapping: Audit major workflows by department (HR: recruiting, onboarding, offboarding; CSM: business reviews, client onboarding; etc.)
Agent inventory: Identify where agents already exist and whether they're truly embedded in processes or used as side tools
Industrialization: Focus on use cases that can be replicated across departments
Depth over breadth: Move beyond company-wide metrics to team-specific penetration

"What's also amazing with these AI projects is that we never know what Dust will release next. There are potentially many game-changing features coming. It's hard to build a long-term roadmap because the landscape will have changed significantly in two months."
— Cécile Hervouet

The vision: AI that disappears

The ultimate goal isn't for employees to "use AI", it's for AI to become invisible infrastructure that simply makes work better.

"What I'd really like to pitch to anyone considering a solution like Dust is that AI in Operations delivers three things: time savings, quality improvement in everything we do, and increased job attractiveness. If we can agentize all the low-value, repetitive, time-consuming tasks, it should increase employee retention and fulfillment." — Cécile Hervouet

Key takeaways for operations leaders

Based on Spendesk's journey from 90% adoption to embedded workflows:

Structure your champions program

Informal enthusiasm isn't enough. Champions need clear roles, dedicated time (even 10%), and accountability to make a real impact.

Measure what matters

Monthly active users are just the start. The real indicators: weekly retention and custom agent usage rates.

Design for retention, not excitement

Hackathons generate buzz; sustained formats generate habits. Consider longer-format initiatives that allow for iteration and feedback.

Interested in learning more about how Dust can help your team? Visit our solutions page or reach out to our sales team.