Exponential Builder
Back to database
Coding & EngineeringBusiness OperationsKnowledge Management

Zach Davis, LaunchDarkly's AI use case

Engineering / product leader at LaunchDarkly

Uses AI agents to make enterprise knowledge and engineering operations more actionable, including centralizing internal docs for agents, identifying tech-debt opportunities, and improving hiring or process workflows.

The problem

What was broken before AI

Large software companies accumulate knowledge in scattered places: docs, Slack threads, code comments, tickets, onboarding material, interview notes, product decisions, and individual memory. Humans learn how to navigate that mess over time, but AI agents struggle when the context is implicit. Without a clearer knowledge layer, agents may miss important constraints, repeat old decisions, or produce work that looks plausible but does not fit the company.

What changed

What the use case made possible

The workflow treats internal context as infrastructure. Docs and process knowledge become material that agents can search, reason over, and use. In engineering, this can help identify tech-debt patterns, repeated friction, or code areas that need cleanup. In operations or hiring, agents can help summarize signals, compare processes, and surface bottlenecks. The key shift is that the company starts shaping its knowledge so agents can participate more reliably.

Why this matters

Why this use case is worth studying

This use case is valuable because it highlights a quiet requirement for enterprise AI: the organization has to become readable. Better agents alone are not enough if the company’s knowledge is scattered and implicit. By making internal context explicit, teams can turn agents from isolated assistants into participants in the operating system of the company.

Use this when

When this pattern applies

Use this pattern when agents keep failing because company context is scattered, outdated, or implicit. It works especially well for enterprise workflows where success depends on knowing internal decisions, systems, process history, and the way work actually gets done.

Exponential Builder analysis

01

Context is infrastructure.

Enterprise agents only become useful when the company’s decisions, docs, code context, and process history are findable and current enough to reason over.

02

Agent failures expose knowledge debt.

When an agent makes a bad assumption, the durable fix is often an updated source of truth, a retired stale doc, or a clearer rule for the next run.

03

Start where people repeat themselves.

Workflows like tech-debt triage, hiring process review, and operational bottleneck analysis are good candidates because they already depend on repeated explanations across scattered systems.

Who this is for

Best fit

Engineering leaders

Developer productivity teams

Operations leaders

Knowledge management teams

Companies adopting internal AI agents

Teams with scattered docs and repeated context questions

Organizations trying to apply agents to tech debt, hiring, or process work

What to avoid

Mistakes and warnings

Where this pattern can go wrong if you copy it too literally.

Do not assume agents can infer company history from scattered docs.

Avoid pointing agents at outdated or conflicting sources without labels.

Do not start with a broad enterprise-wide agent before one workflow works.

Keep humans responsible for decisions that affect hiring, code quality, or team process.

Treat agent failures as signals that the knowledge layer needs work.

Public workflow preview

The shape of the workflow

A high-level look at how the use case works, with the reusable pattern made clear.

01

Find where company context is scattered

Start with docs, tickets, code, Slack, onboarding notes, and process materials.

02

Make the context agent-readable

Organize the information so AI can find the right source, understand the decision history, and avoid outdated assumptions.

03

Point agents at repeated operational problems

Use cases like tech debt, hiring workflows, and process bottlenecks become easier when the agent can see the supporting context.

04

Keep humans in the review loop

Agents can summarize, flag, and propose, but teams decide what is correct and worth acting on.

05

Turn every run into better context

When an agent gets stuck or misses something, the fix becomes a doc, rule, or source update for the next run.

Copy the pattern

The reusable idea

Pattern in one sentence

Make the company readable to agents by turning scattered internal context into source-of-truth material they can use and improve.

Reusable idea

Zach’s use case is a reminder that agent adoption is partly a documentation problem. If an agent keeps asking for context, making the wrong assumption, or missing company-specific constraints, the answer may not be a better prompt. It may be a better internal knowledge layer. The more clearly a company explains itself, the more useful agents become.

Steal this workflow

Run a one-workflow “agent readability” pass:

1

Pick one workflow where people constantly re-explain context, such as tech-debt review, hiring process analysis, or engineering handoffs.

2

List every source the agent would need: docs, tickets, code notes, Slack threads, onboarding material, interview notes, or process pages.

3

Mark each source as current, stale, duplicated, conflicting, or missing.

4

Create one source-of-truth map with: key docs, decision history, known constraints, owner, last-updated date, and sources the agent should avoid.

5

Give the agent one narrow job, such as “group repeated engineering friction by cleanup opportunity.”

6

Have the domain owner review the output for wrong assumptions, missing constraints, and useful findings.

7

Convert every miss into a doc update, rule, label, or deleted stale source.

8

Repeat until the agent needs less clarification, then expand to the next workflow.

Suggested prompt

“You are helping us make this workflow agent-readable. Review the provided docs, tickets, notes, and process materials. Identify: 1) which sources appear current, stale, duplicated, conflicting, or missing; 2) what context an AI agent would need before doing useful work on this workflow; 3) the most likely assumptions an agent would make incorrectly; 4) the smallest documentation or rule updates that would prevent those mistakes; and 5) one narrow agent task we can safely test next with human review.”

Field notes

Get new AI use cases in your inbox

A short weekly note on how real people are using AI to save time, make money, build tools, and run their lives.

No spam. Just useful AI use cases.

Related use cases

Keep exploring nearby systems.

Browse all
Minerva CQ team headshot
Customer SupportSales & Outreach+1

Minerva CQ team

AI Research Division / product team at Minerva CQ

Real-time contact-center co-pilot workflow where AI listens to live voice support, detects intent and sentiment, retrieves context, updates customer profiles, and guides agents through next-best actions.

Minerva CQOpenSearchLLM APIs
Use case
View use case
Business OperationsCompany AI Adoption

Company S expense processing team

Finance operations team at Company S

Corporate expense-processing workflow where receipts are recognized with OCR/IDP, classified against policy data, routed through generative-AI exception handling, and finalized with human review.

Brity AutomationBrity Automation AI FlowSAP
Use case
View use case
Included Health clinical team headshot
Business OperationsKnowledge Management+1

Included Health clinical team

Clinical team using an internal ambient scribe workflow at Included Health

Telehealth ambient scribe workflow where visit audio is transcribed with Whisper and processed with GPT-4o into SOAP notes and patient instructions for clinician review.

WhisperChatGPTOpenAI Audio API+5
Use case
View use case