back_office_ops · saas · workflow
How Spendesk turned 90% adoption into embedded AI workflows with a champions program
After achieving 90% AI adoption, Spendesk found that depth of usage was uneven — most employees were using Dust as a generic chat replacement rather than in embedded workflows, and an initial limited rollout had created a perception that AI was only for certain roles.
How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · Company-wide platform deployment
The Dust platform expanded from a limited pilot to the entire company by May 2025.
Tools used
Dust
Outcome
By December 2025, Spendesk achieved 93-94% monthly active users and 80%+ weekly retention, with 40%+ of messages going to purpose-built custom agents rather than generic LLM interactions.
What failed first
A summer hackathon that generated 11 new agents resulted in only 1 surviving after six months, with momentum evaporating almost immediately — demonstrating that burst-format events do not produce sustainable AI habits.
Results
Time saved93-94%
Volume83%+
Running sinceFebruary 2025
Grounding & classification
Source type: vendor customer story
36 fields verified against source quotes.
agentic workflowdata extractionenterprise searchknowledge searchknowledge basefailure mode describedmetric backednamed customerproduction runtime claimedtools describedvendor confirmedworkflow describedfinancial servicessoftwareautomation rateemployee productivitytime savedvendor customer storyback office opshr opssales opsagentic task executionrag answering