_1771094022787-vmkQmLLT.png)
Define the job.
Use Startedby’s discovery and company context to understand the people, systems, steps, and exceptions involved. Refine the improvement into a specification: the outcome, trigger, required information, and human decisions.
Custom Harness
A purpose-built environment
for the work that needs one.
Some workflows need their own place: a clear starting point, the right information, and a deliberate path from request to review.
A Custom Harness brings AI and the environment around it together for that work. It can include a traditional web application, with screens, buttons, structured records, and client access. Its design also covers the company context, instructions, tools, permissions, and controls that make AI useful within the workflow.
Talk with an expertTuesday, 18 June
Good morning, Jordan
Operating picture
A live view of where your client work needs attention.
Active clients
08
+2 this quarter
Open decisions
14
3 due this week
Sessions logged
126
92% summarized
Context health
Signal quality
Company context86%
Decision coverage68%
Recent evidence92%
Delivery pulse
On track this week
MonToday
Priority queue
Work that keeps momentum
NorthstarConfirm delivery owner for the launch workflowDue today
Cedar & Co.Review the latest findings synthesisTomorrow
Build around a real need.
Most companies can start with an AI environment their people already use. A Custom Harness makes sense when a particular workflow needs more specific structure.
Your team may need to review reports in a queue, give clients access to selected information, or keep a consistent record of requests and decisions. People should be able to see what needs attention and what happens next.
An existing application may provide part of that environment. The scope can include extending it with AI and the surrounding controls, where its capabilities and access allow.
The application is part of the harness.
A web application supplies the screens and software behavior. The complete harness connects those with the AI model or agent and the conditions it needs to work well.
Company understanding
The work draws on relevant Startedby context: responsibilities, workflows, conventions, handoffs, and exceptions. Decide what the environment needs to retrieve and what each person is allowed to see.
AI, instructions, and tools
Define where AI helps, what outcome it should work toward, which tools and data it can use, and what it should do when information is missing or conflicting. Some steps may be ordinary software; others may call for AI reasoning.
Permissions and human review
Define who can request work, review a result, approve a next step, or change information. The implementation needs explicit access and controls for each responsibility, including any operation that changes another system.
An interface that fits the job
Design the screens, forms, buttons, and records around how people actually work. A client portal or internal review screen can sit alongside an AI agent operating within the same purpose-built environment.
A useful Custom Harness makes these parts work together. A polished screen alone does not give AI the context or authority to do the job.
From report request to reviewed result.
Consider a team preparing recurring client reports. An existing AI tool may be a good place to request and review a draft.
A Custom Harness could instead give the team a report queue: select a client and period, prepare a draft using permitted information, examine missing details, and route the result to the responsible reviewer.
Behind those screens, the AI needs the reporting instructions and relevant company context. The workflow needs configured data access and a clear distinction between preparing, approving, and delivering the report.
Those are design choices to implement and test together. A button should make the next step clear, including when a person needs to decide.
Understand the work before you build.
Scope the environment.
Work with an expert to decide which parts belong in existing tools and which need a purpose-built experience. Include the AI behavior, context access, data connections, permissions, review steps, interface, and responsibility for ongoing maintenance.
Build and check it against real work.
The implementation partner builds the agreed environment and tests it with representative tasks and exceptions. Your team checks the results and decides what can be used, by whom, and under which controls.
Startedby provides the discovery and company-context foundation. Custom Harness design and development are separately scoped expert services; a Startedby seat does not include a custom build.
Keep the understanding connected.
A Custom Harness can use Startedby’s permitted context through MCP, or use reports and Markdown exports where appropriate. Live access and exported snapshots have different update needs.
Data connections and operations must be scoped for the particular implementation. Startedby’s currently implemented system connection is HubSpot. Reading information through MCP does not authorize write access to another system.
As the company’s work changes, review the harness’s instructions, access, and behavior alongside the company understanding. Ongoing ownership is part of the design.
Choose the right amount of structure.
Use an existing harness when it fits. Build a Custom Harness when the work benefits from its own environment.
Back to Harness to compare the two paths.
Learn about Experts and agree on the scope and fees directly with your expert. Services are separate from your Startedby subscription.
Talk with an expertSee For experts.
Build when the work deserves its own environment.
Start with the work, then choose the right amount of structure.