Knowledge Agent

From a call to Salesforce

One conversation, three agents, nobody retyping anything. A call is transcribed, the Knowledge Agent turns what resolved it into a reusable article, and the result is written back into Salesforce — a concrete example of agents handing work between them rather than being separate features you switch on individually.

  • A call becomes a reusable article without anyone writing it
  • Three agents on one conversation, in sequence
  • The result lands back in Salesforce
  • Knowledge captured at the moment it was created

Watch the full tour

Verify your work email once and every research paper, case study and product tour on ascendo.ai opens, right here, without leaving the page.

  • Takes about 30 seconds
  • Work email only, no spam

The best answer your company produced today was spoken aloud and lost

Somewhere this week an engineer explained on a call exactly how to fix something difficult. That explanation was the most valuable knowledge your company produced that day, and it existed for four minutes.

Not because nobody cares — because writing it up is a separate task afterwards for a person with another call waiting. And when somebody does, it lands in a knowledge tool the CRM knows nothing about.

Three cards tracking an engineer’s spoken explanation: four minutes of the most valuable knowledge produced that day, never written up because the queue wins, and when it is, filed where the CRM cannot see it.
Even the good outcome fails at the third card, because the person handling that account next works in the CRM.

The chain

Four steps, three agents, and no point where a person has to remember to do something.

  1. Step 1

    The call is transcribed

    The conversation becomes text automatically, so the raw material for everything downstream exists without anybody producing it.

  2. Step 2

    The resolution is identified

    What actually resolved the problem is picked out. Most of a call is not the answer.

  3. Step 3

    An article is drafted

    The Knowledge Agent turns that into an article structured the way your knowledge base expects — a draft for review, not a publication.

  4. Step 4

    The result reaches Salesforce

    What was learned is written back into the record, so the next person on that account sees it.

Four stages: the call transcribed automatically, the resolution picked out of the conversation, an article drafted in the knowledge base’s shape, and the result written back into the Salesforce record.
No stage here waits for somebody to remember it, which is the only reason the chain survives a busy afternoon.

Why the chain matters more than any one step

Each of these steps alone would leave a manual handoff at the join.

Capture happens while the knowledge exists

The window is short. An explanation is complete and precise while the person giving it still has the problem in their head.

Three agents, one piece of work

Transcription hands to knowledge generation, which hands to the CRM write-back, and the context survives all three. Any one alone leaves somebody copying between tools.

A person still reviews it

A generated draft is a starting point, not a publication. Somebody checks it before it goes live — five minutes on a draft, an hour on a blank page.

A six-row table naming each step of the chain, which agent performs it, and what a person would have to do by hand if that step were missing.
Read the right-hand column on its own: it is the list of small manual jobs that quietly stop happening in a busy week.

Why the usual approaches fall short

Knowledge capture competes with the queue, and the queue always wins.

Asking agents to write up their calls

Competes directly with the queue and loses. Coverage is worst on the hard problems.

Recording calls and keeping the audio

That is storage, not knowledge. Nobody searches a recording when they need the answer.

A quarterly knowledge sprint

Written from memory months later. Accurate in outline, wrong in the details that mattered.

A knowledge base disconnected from the CRM

The article exists, and the person working the account never finds out that it does.

Two columns comparing asking agents to write up calls, keeping recordings, quarterly knowledge sprints and a knowledge base disconnected from the CRM with an automatic chain.
Keeping the audio is the one that feels like progress: the knowledge is technically retained and nobody will ever search a recording for it.

What things are called

The three handoffs in this chain, named.

Transcript
The text of the conversation, produced automatically as raw material for everything after it.
Knowledge Agent
The agent that turns resolved work into structured, reusable articles.
Write-back
Returning the result into the system of record, so it sits where the next person will actually look.
Agent handoff
Passing work between agents with the context intact rather than restarting it.
Draft
A generated article awaiting human review. Not a publication.

Frequently Asked Questions

No. It is a draft for a person to review for accuracy, tone and clarity before publication. That is deliberate: drafting is the slow part and judgement is the part that has to stay human.

Where this fits

Knowledge from Slack

The same capture pattern in a text channel: a resolved thread becoming a reviewable article.

AI Companion for Salesforce

What agents see inside Salesforce once this knowledge exists to be surfaced.

Knowledge Intelligence Agent

The agent doing the drafting, and how articles are structured and kept current.

More Knowledge Agent tours

All tours

Knowledge Agent

Capture the calls you already record

If you record support calls, the raw material is already sitting there. Send us a handful and we will show you the articles that would have come out of them.

Talk to us