back_office_ops · ecommerce · workflow
Instacart builds Ava, a company-wide internal AI assistant powered by GPT-4
Instacart had no company-wide AI tool despite strong productivity gains during a hackathon with ChatGPT; non-engineering employees also faced a blank-text-box barrier that prevented broader AI adoption.
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 · User submits request
Users initiate tasks with Ava to write, review and debug code, improve communications, or learn faster.
Tools used
AvaGPT-4GPT-3.5ChatGPT
Outcome
Ava reached company-wide adoption with over half of Instacart employees using it monthly and more than 900 weekly; the Slack thread summarization reduced a five-to-ten minute context grind to just a few seconds of skimming.
Results
Time savedover half of Instacart employees
Volumemore than 900
Grounding & classification
Source type: technical build writeup
28 fields verified against source quotes.
chatbotcode generationconversational aisummarizationchat transcripthuman review describedmetric backednamed customerproduction runtime claimedtools describedworkflow describedecommerceemployee productivitythroughput increasetime savedtechnical build writeupback office opsit supportai draft human approvalmeeting to artifacts