Shahzad Ali · Learning notes
Example workflows for acquisition research and sales research.
A coordinator can divide a broad question into focused research tasks, then combine the findings into something a person can use.
Two examples of moving from a clear brief to something a person can use.
Should we investigate this company further?
Revenue, margins, cash and assumptions.
Customers, competitors and demand.
Obligations, uncertainties and questions.
Combine sources, reconcile assumptions and surface disagreements.
A person evaluates the case and chooses the next step.
Which companies fit our target customer?
Search against the agreed criteria.
Verify company facts and relevant contacts.
Prepare outreach grounded in that evidence.
Check fit, source dates, accuracy and message quality.
External actions follow the authority given to the workflow.
Finance example
One workstream can inspect financial statements, another market evidence, and another contract terms. The coordinating agent can bring their sources and assumptions together.
A valuation depends on verified inputs, a stated method and sensitivity to assumptions.
How tasks become tool useSynthesis
The final pass should compare dates, definitions, assumptions and contradictory findings. A clear report distinguishes sourced facts from estimates and unresolved questions.
A polished report can still contain errors. I want the evidence trail to survive the handoff between agents.
How to check a systemSales example
An agent can organise public company information against stated criteria and prepare outreach drafts. A connected CRM can hold the reviewed records.
Useful prospect research records the source, date and reason each company meets the brief.
Retrieving relevant informationExternal actions
A workflow can stop with drafts for review, or request approval for a defined action. Sending messages and updating live records require the corresponding authority.
Decide whether the task should end with a draft, an approved update or an authorised message.
The role of toolsSources behind the technical details
Reviewed . Technical details are linked to their sources. My perspective on how I use AI is personal.
Anthropic. Engineering patterns for workflows, autonomous loops and orchestrator-worker systems.
Anthropic. Describes parallel research, coordination, evaluation and resource costs in one deployed research system.
OpenAI. Shows how applications pause tool execution for approval and resume from saved state.