Spendesk achieves 90% company-wide AI adoption in 6 months with Dust
The problem
Spendesk employees wanted AI tools to boost productivity but security and compliance requirements blocked unrestricted use of external AI, while point solutions explored for specific use cases would create vendor lock-in per use case rather than a company-wide benefit.
First attempt
Point solutions evaluated for customer support and sales acceleration were rejected because they would lock Spendesk into a separate vendor for each individual use case instead of providing a single platform that could benefit all teams.
Workflow diagram · grounded in source
1
Three-month customer support POC
Trigger
Spendesk launched a three-month POC focusing on customer support to begin AI experimentation.
▾ source quote
“Spendesk launched a three-month POC focusing on customer support”
2
Leadership review of qualitative feedback
Human review
Leadership evaluated qualitative employee feedback rather than hard metrics to assess readiness for full deployment.
▾ source quote
“At that time I didn't have metrics to share in terms of 'we saved 1,000 or 1,500 hours,' but when people told me, 'Greyg, tomorrow I can't imagine working without Dust,' that was strong measurable feedback”
3
Full company-wide deployment
Integration
By January 2025, leadership decided to move forward with full deployment, expanding scope to RFP response tools and company-wide knowledge agents.
▾ source quote
“By January 2025, after seeing strong results, leadership decided to move forward with full deployment. The scope quickly expanded beyond support to include RFP response tools and company-wide knowledge agents.”
4
Hackathons for agent experimentation
Feedback loop
Teams competed in hackathons to create the best Dust agents for their workflows, generating enthusiasm and practical use cases.
▾ source quote
“Spendesk launched a series of hackathons to accelerate learning and experimentation where teams competed to create the best Dust agents for their workflows, generating enthusiasm and practical use cases”
5
AI Champions program across departments
Routing
One or two AI champions were placed in each department to understand team needs and implement AI into workflows.
▾ source quote
“launched an AI Champions program, which placed one or two champions in each department: individuals responsible for understanding their team's needs and implementing AI into workflows”
6
Employee-built custom agents
Output
Employees build custom Dust agents for their teams, with 1 in 4 users becoming agent builders.
▾ source quote
“These agents are built by Spendeskers for Spendeskers to ease existing operations, 1 in 4 users are agent builders”
Reported outcome
Spendesk achieved 90% company-wide AI adoption within 6 months, reached 92% weekly user retention, and had 1 in 4 users building custom agents, while replacing multiple point solutions and eliminating redundant AI subscriptions.
Reported metrics
company-wide AI adoption rate90%
Weekly user retention92%
Users who are agent builders1 in 4
messages sent to custom Dust agentsaround 50%
Show all 5 reported metrics
company-wide AI adoption rate90%
weekly user retention92%
users who are agent builders1 in 4
messages sent to custom Dust agentsaround 50%
redundant AI subscriptions eliminatedelimination of redundant AI subscriptions and specialized vertical solutions
Reported stack
Dust
◆ Does this fit your context?
Compare to your context
Tell us your scale, team, and constraints. We'll show what changes at your size, what fails at your scale, and whether this case is a fit, needs adaptation, or won't scale to you. Free demo, no signup.
Spendesk achieved 90% company-wide AI adoption within 6 months, reached 92% weekly user retention, and had 1 in 4 users building custom agents, while replacing multiple point solutions and eliminating redundant AI sub…
What tools did this team use?
Dust.
What results were reported?
company-wide AI adoption rate: 90%; Weekly user retention: 92%; Users who are agent builders: 1 in 4; messages sent to custom Dust agents: around 50% (source-reported, not independently verified).
What failed first in this deployment?
Point solutions evaluated for customer support and sales acceleration were rejected because they would lock Spendesk into a separate vendor for each individual use case instead of providing a single platform that coul…
How is this back office ops AI workflow structured?
Three-month customer support POC → Leadership review of qualitative feedback → Full company-wide deployment → Hackathons for agent experimentation → AI Champions program across departments → Employee-built custom agents.
This case is one data point. Whether its pattern fits you depends on your volumes, your stack, and your exception load — that comparison is the step no case study can do for you.