AMove
Enterprise & Business Software · File & Knowledge Management
From reliability foundations to AI knowledge retrieval inside a live cloud product
A longitudinal brownfield engagement: reliability and coverage work first, then AI-assisted ingestion and retrieval across knowledge bases held in different cloud storage services.
The starting point
A file and cloud management product, desktop application and web portal, in use by customers across multiple cloud storage services.
What had to change
Critical paths built up over years had uneven coverage, and later the product needed knowledge scattered across different cloud storage to be retrievable inside the workflow.
What couldn't break
Everything had to be delivered without freezing the roadmap or disrupting a product customers depend on.
Why this was hard
Coverage had to be earned across a multi-year codebase, and knowledge retrieval had to work across knowledge bases stored in different cloud services with different access models.
What Datics changed
First a structured quality and regression strategy across the portal's critical paths, then AI-assisted ingestion, vectorization and retrieval across knowledge bases in different cloud storage, with Slack interaction and workflow automation.
In sequence, alongside ongoing product work: stabilize the foundation, then add the AI retrieval and workflow capability on top of it.
Reliability foundation
A structured quality and regression strategy across the portal's critical paths.
AI ingestion and vectorization
AI-assisted ingestion and vectorization of knowledge bases.
Cross-cloud retrieval
Retrieval across knowledge held in different cloud storage services.
Slack workflow automation
Slack interaction and workflow automation on top of retrieval.
What changed
A more predictable product, and knowledge held across different cloud storage services now retrievable through the workflows people already use.
- Delivery freeze required
- None. Work added alongside feature delivery
- Phase one
- Regression coverage across a multi-year codebase
- Phase two
- AI ingestion, vectorization and retrieval
Under the hood
- Datics role
- Quality engineering followed by AI product engineering: test strategy and regression coverage, then ingestion, vectorization, retrieval and Slack workflow automation.
- Integrations
- The product's desktop application, web portal, the cloud storage services behind them, and Slack.
- Deployment
- Customer's production cloud environment on AWS, with OpenAI, Python, Node, React, LangChain, a vector database and the Slack API; tested with Selenium, Cypress, JMeter and OWASP ZAP.
- Production boundary
- Customer's production cloud environment.
Technology
What this unlocked
The retrieval and automation layer is the base for further capability inside the product.
Related work
TEG
Enterprise & Business Software · Business Process / BPMN Modeling
Evolving an AI BPMN modeling suite through six capability phases
One engagement spanning six releases of an AI-assisted BPMN enterprise modeling suite, with capability added phase by phase rather than in a single build.
COSMIC Platform
Enterprise & Business Software · KPI / Analytics SaaS
Conversational business intelligence for SME operating metrics
COSMIC Platform is a B2B business-intelligence and KPI analytics system designed to turn operational data into quicker interpretation and AI-assisted decision support.
Peerlogic
Dental & Veterinary · Practice Communications SaaS
AI communication intelligence inside a live practice platform
Established communication intelligence software serving dental and veterinary practices. AI needed to become part of existing practice workflows rather than a separate application.
Have a product with the same kind of constraints?
Tell us what already exists and what you want it to do. We will tell you what it takes.