TELOA BLOG

Why Teloa has both conversation and work interfaces

Intent develops through discussion; decisions need actual content. Teloa connects conversation, inspection, and artifact revision so work can improve after the first result.

People often begin work without knowing exactly what the delivery should be. “Help me look through these materials” may become a decision about service coverage rather than price once the comparison is available. Requirements form through collaboration; a product can rarely ask for all of them before the first sentence.

Yet when someone decides to change a price, accept a report, or authorize execution, a rough understanding in conversation is no longer enough. They need to see the exact object and content, and understand what the decision will change.

Together, these needs shape Teloa’s interaction. Conversation accommodates intent that is still forming; work interfaces handle concrete content and actions. Our concern is how people move between them and gradually get the work right.

While intent is forming, let people explain the problem

Consider a hypothetical supplier comparison. The user initially wants to “choose a suitable supplier,” then adds a budget, support requirements, and a delivery date. Requiring every field at the start may collect premature decisions made just to get through a form.

Conversation lets people add information, ask questions, and reconsider an initial assumption. The main assistant can help clarify the problem first. Work that needs sustained execution can then be explicitly delegated, and a recurring responsibility may later justify an employee. Structure should be introduced as the work becomes clear.

The squirrel’s small window and full main chat share a conversation for the same reason. A brief question can be asked beside a page; a longer discussion can move into a larger space while retaining the requirements already formed. This reduces the cost of starting and continuing. It does not imply that the assistant has read everything on the current screen.

Once content is concrete, decisions need to be concrete too

The comparison might reveal that the quotes cover different services. A request to “make the conditions clear” still requires checking what was added. The model saying “updated” cannot replace that inspection.

Here an interface offers precision that conversation handles less well. Text can be read and edited directly, business changes can show field differences, and an execution approval can present an explicit choice. Deciding on the actual content is more reliable than inferring it from an action description.

Community’s supported business-change path saves a draft first. The owner then requests adoption, with approval and execution handled through the original conversation. Natural-language business building also begins with a supported configuration draft. The inspection is intentional: removing manual entry still needs to leave the system’s interpretation visible.

Precise operations add interface and version-management costs. But letting the model write directly from its interpretation makes the user discover afterward which date, customer, or condition it misunderstood. Putting the check at the decision point is a product cost we choose to carry.

After the first draft, collaboration becomes better informed

Many requirements become expressible only after a first draft exists. Once the supplier table is available, a person may keep the price differences but ask for excluded services to be listed separately. That feedback is more specific than the original request because both sides now have visible content to discuss.

An artifact therefore needs to be accessible, inspectable, and revisable. Leaving long text in a reply makes it harder to identify the content a person accepted or determine what a later turn changed.

The current general-artifact path saves feedback against an explicit version. After the owner hands it back and approves execution, a follow-up task can write a new version of the same artifact. The original task retains its accepted content. Continuing a revision becomes traceable collaboration rather than the generation of another similar text.

For the user, the round trip is simple:

  1. Express a goal in natural language before every detail is settled.
  2. Inspect the actual content and identify differences or change a decision.
  3. Obtain the revision through the original work and assess whether it is useful.

The product’s structure exists to connect those three moments.

A change should remain valid when the form of interaction changes

The same service description may be drafted in conversation, have its conditions edited in the library, and then return to chat for examples. The third operation needs the content saved by the second. Otherwise, each change of entry point requires the user to synchronize their work again.

Teloa’s supported natural-language resource operations reuse the original services and preserve object identity, version, and scope. Conversation becomes a maintenance method while the original interface remains useful. Ambiguous references and newer content require checks against the actual object, handled by the shared service.

Following the process needs its own division of work. The work overview places progress, background work, and artifacts nearby to help a person decide whether to wait or inspect. Temporary details can be read and closed. These interfaces provide necessary evidence beside the conversation without requiring a full reading of execution logs.

Leave room for a change of mind

Conversation alone can bury concrete decisions in exchanges. Fixed interfaces alone make unsettled requirements difficult to express. Working together requires accommodating both limits and connecting the two accurately.

We want people to be able to start earlier and change their minds later. The material, inspection, and judgment invested in one round should help the next get closer to the result. That extends the measure of interaction from how easy the first input was to how well the work can improve after someone has seen it.

Original
1.0×

Double-tap or pinch to zoom. Drag to explore.

Keep reading

All articles
Product thinking

From individual agents to an AI team

As AI moves from executing individual tasks to holding lasting responsibilities, a product has to retain goals, working relationships, and what the team builds together.

Product thinking

Execution ends. How does experience remain?

Work records need accuracy; experience needs room for reinterpretation. Teloa maintains both time scales and selects valid experience for later work.