back office ops · pattern

Internal AI copilots

Org-wide AI assistants — Dust-style — adopted by teams to query systems, draft, and automate routine work.

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 · Connector & data access setup

The copilot is wired to company systems (CRM, ticketing, docs, calendar, repos) with permission scopes that mirror each employee's existing access — no new permission surface.

What fails first / common problems

Recurring first-deployment failures from the matching workflows'what_failednotes. First sentence of each, attributed to the source case.

Individual automation setups using Relay, Zapier, or personal Claude MCP configurations did not scale because each workflow was tied to a single employee's account and required technical setup most staff could not do.

from: Assembled builds a company-wide AI operating system with Dust, achieving 95% internal adoption across 120+ employees

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.

from: Spendesk achieves 90% company-wide AI adoption in 6 months with Dust

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.

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

Generic AI assistants and existing Text-to-SQL tools were found unsuitable: generic AI tools lack business-specific context, and text-to-SQL products are designed for non-technical business users rather than engineers.

from: BlaBlaCar builds a Data Copilot to shift data analysis left with LLMs and GitHub Copilot

Existing self-service tools were sub-optimal because they assumed users already knew which data sources to query and how to interpret them correctly; skillset gaps and the risk of misinterpretation limited their usefulness for critical a…

from: DoorDash builds a multi-agent internal AI platform for unified data and operations intelligence

Tools commonly seen

dustgithub copilotcursorslackavachatgptgpt-4a2aairflowbigqueryblablacar data copilotbm25

Representative outcomes

Real metrics from selected cases — verbatim from each workflow'snumberspanel. Click any title to open the full case.

Assembled builds a company-wide AI operating system with Dust, achieving 95% internal adoption across 120+ employees

Time savedhundreds of hours

Volume95%

Spendesk achieves 90% company-wide AI adoption in 6 months with Dust

Time saved92%

Volume90%

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