Agents | Field note

Agents for real workflows

An agent is only interesting when it can survive the boring parts of work: context, repetition, exceptions and handoffs.

The seductive version of agents is autonomy. The useful version is often discipline.

Real work is full of partial context, stale assumptions, approvals, handoffs and small decisions that are easy to describe in a workshop and harder to support in practice. Agent prototypes make those gaps visible quickly.

What the work teaches

The strongest agent concepts do not begin with "what can the agent do?" They begin with "where does the work keep losing shape?" That might be between systems, between people, between meetings, or between a decision and the evidence that should support it.

The prototype has to expose the boundary of delegation. What should be automated? What should be prepared? What should be escalated? What should remain stubbornly human?

Reusable lesson

Design the handoff before the autonomy. A good agent leaves a trail: what it used, what it inferred, what it skipped and where human judgment must enter.

That makes the work less spectacular but more usable. The goal is not to remove people from the loop. It is to make the loop worth being in.

Concrete lessons

LESSONS

Shape the workflow

Start with the work before the agent

The strongest agent work does not begin with autonomy. It begins with understanding the workflow: who will use the output, what decisions it should support, what data it can see, what uncertainty matters and where the handoff to a human or team happens.

User research is not separate from agent design

Qualitative interviews with future users can create the foundation before the agent itself is built. Interviews expose the real job: what users trust, what they need explained, what would slow them down and what kind of output a technical team can actually build from.

Design the handoff before the autonomy

A useful agent should leave a trail: what it used, what it inferred, what it skipped and what needs human judgment. The goal is not to make the agent look independent. The goal is to make the next human decision easier.

Separate user modes from system capability

Not every user should get the same output. A standard user may need a briefing. A steward or specialist may need review triggers, source gaps or escalation material. Good agents define who can receive what, instead of exposing the whole machine to everyone.

Control the evidence

Keep baseline and live signals visibly separate

For example: a frozen knowledge baseline is not the same thing as a current signal. A live web source can strengthen monitoring, but it should not silently rewrite the canonical picture.

Read-only is a product decision, not just a permission setting

A read-only agent should say what it is allowed to do and what it must never do. It can compare, brief, flag and recommend review. It should not mutate canonical data, change scores or pretend that a baseline has been updated.

Route to the smallest useful source set

Large agents need source routing. The agent should know which baseline layers, source baskets and overlays matter for the question. Reading everything is not thoroughness; it is often noise wearing a serious coat.

Dates are part of the evidence

For agents working with current signals, recency must be explicit. A source without a date should not carry a time-sensitive conclusion. "Latest" is not a vibe; it is a retrieval and evidence problem.

Keep judgment honest

Classify the source basis before analyzing

For example: full text, partial text, metadata, secondary summary and mismatched source are different evidence situations. An agent should downgrade or stop when the source basis is too weak.

Confidence and overreach must be visible

Agents become dangerous when they sound certain on thin evidence. Good outputs show confidence, overreach risk, alternative readings and what cannot be inferred from the available material.

Do not confuse signal with conclusion

One news item is not automatically a new risk. One biased text is not automatically propaganda. One weak signal is not enough for a strong claim. The agent should compare, qualify and mark uncertainty before it escalates.

Use comparison without false certainty

A source mirror can help users understand whether a text's framing is typical or distinctive. But comparison is not the same as fact-checking, and it should not create false balance or override the primary analysis.

Build the agent as a system

Skills can become agent modules

A useful pattern is to let the agent own the role, boundary and workflow while a bundled skill owns the reusable method. That makes the system easier to improve: the surface can change without burying the analytical craft.

Test the behaviors the agent must refuse

Agent QA should include forbidden requests: change the score, overwrite the baseline, infer intent from metadata, create a new risk from one incident. The best tests are not only "can it answer?" but "does it stop in the right place?"