Discovery before development
Five steps. Discovery is the first three. None of them commits you to the next.
-
Strategy Call
No cost. We learn the problem, the workflow, the department, the systems involved and what you want to achieve.
-
Initial Discovery
One or two additional conversations at no cost, when the first call shows there is something worth examining.
-
Deep-Dive Discovery, $2,000
Paid workflow analysis and technical diligence. Credited in full against the Statement of Work if the project proceeds.
-
Build
Scope, deliverables and architecture agreed, then built.
-
Run
Operated and improved on HeyQuinn or platform-native infrastructure.
The 15 questions
Answered before any technology decision is made.
- How is the work performed today?
- Where does information originate?
- Who makes decisions?
- What systems are involved?
- Where are employees manually moving information between systems?
- Which decisions are deterministic?
- Which decisions require judgment?
- Where can AI safely make or recommend decisions?
- What happens when AI is uncertain?
- What permissions should an agent have?
- What systems should it be permitted to access?
- What actions should require human approval?
- How will the system be monitored?
- How will its actions be audited?
- How should the workflow evolve over time?
The architecture lens
Every workflow is examined through eight layers. A gap in any one of them shows up in production.
Technical concepts, explained operationally
- AI agent
- Software that can take multi-step action toward a goal, rather than only answering a question.
- Agent orchestration
- Several specialized agents handing work to each other, each responsible for one part of a workflow.
- MCP (Model Context Protocol)
- A standardized, controlled way for AI agents to work with business systems: retrieve information, use approved tools and take defined actions.
- Model routing
- Different steps of a workflow can use different models, chosen for cost, performance or data sensitivity.
- Tools & skills
- The defined, approved actions an agent is allowed to perform. Nothing more.
- Human-in-the-loop architecture
- Points in a workflow where a person must review or approve before the agent proceeds.
- Least-privilege access
- The systems and data an agent can reach, scoped as narrowly as the work allows.
- Observability
- Visibility into what an agent did, when and why, both in real time and after the fact.
- AI runtime
- The operating layer between AI models and the business: the tools, permissions, workflows, integrations, monitoring and controls required to put AI to work.
Frequently asked questions
What happens during discovery?
Discovery maps how the work happens today, including the systems, the people and the exceptions, before any technology decision is made.
Why is Deep-Dive Discovery $2,000?
It pays for real analysis: process mapping, systems review, a security and permissions assessment, and a preliminary architecture. It is not a sales conversation.
What happens to the $2,000 if we proceed?
It is credited in full against the Statement of Work.
How do you decide which workflows should use AI?
We ask where decisions are deterministic, where they require judgment, and where AI can safely act or recommend. We do not start from what is technically possible.
What do we get at the end of Deep-Dive Discovery?
The analysis behind the $2,000: a map of the workflow as it runs today, a systems review, a security and permissions assessment, and a preliminary architecture. If the project proceeds, that work becomes the basis of the Statement of Work.
Do we have to proceed after Deep-Dive Discovery?
No. If you proceed, the $2,000 is credited in full against the Statement of Work. If you don't, the analysis is yours to keep.