AI agents

AI agents that do the HR work, inside limits you set.

They draft plans, training and reviews from your documents, check what people understood, and take actions when asked, always with the asking person's permissions.

What you get

Real work, with a person accountable.

01

Each agent has one job

Drafting agents write first versions, examiners mark answers, an insights agent finds skill gaps and overdue work, and the assistant takes requests.

02

Borrowed authority only

An agent acts with the asking person's own sign-in. There is no agent account, so it cannot do anything that person could not do by hand.

03

A dial per area

Every area starts at "proposes only". Admins can let agents do more as trust grows, and anything over 10 people always waits.

How it works

Three settings, chosen per area.

1

Proposes only

The default. The agent prepares the action and nothing changes until a person confirms.

2

Asks first on risky ones

Actions that can be undone run; anything irreversible waits for a yes.

3

Acts, tells you

Available per area, except performance, succession and offboarding, which the database refuses to run unattended.

The difference

How agent platforms differ.

HR chatbotsGeneral agent platformsAscend
Takes actionsNo, it answers questionsAcross whatever tools it is givenYes, inside HR and learning work
Whose permissionsVariesOften a shared service accountThe asking person's own sign-in
Decisions about peopleNot applicableDepends on set-upNever approves or signs off for anyone
Audit trailConversation historyOften tool-call logsAn append-only log naming the person and the agent
Questions

What buyers ask.

What can the agents never do?
Approve, sign off or attest for anyone. Publish training a person has not approved block by block. Draft someone's goals. Act on more than 10 people without confirmation. Run unattended in performance, succession or offboarding.
Who is accountable for what an agent does?
The person who asked for it or confirmed it. The audit log names that person on every row, with the agent recorded beside them.
Which model provider do the agents use?
Language model calls go to Anthropic. Document search, reranking and narration audio run on models Ascend hosts itself.
Can we turn it off?
Leave every area at "proposes only" and the agents can prepare actions but never carry one out without a person confirming. Questions and lookups still work.

See it working on one of your own job roles.

A 30-minute call. Send two or three documents for one role, and watch Ascend turn them into an onboarding plan, training and a check that people understood it.