AI teammates

Build a team. Not another chat.

Bring the customer question into a room. Ask Riley to check the policy, then Jamie to prepare the handoff. Your people check the evidence and decide what happens next.

Join the waitlistSee it in practice

Mac-only for now. Free during the beta (invite-only).

Explore a team

Different jobs. One piece of work.

Your people direct the teammates and make the final call.

Support deskRoom

A billing question needs more than a quick reply.

Start withTicket notes + billing records + approved policy

  1. RileyCustomer support

    Prepare the reply. Flag the payment check.

    Gemini
  2. SageTeam knowledge

    Find the policy and the reason behind it.

    Claude
  3. JamieCustomer handoffs

    Prepare a handoff for the billing team.

    OpenAI

Shared company memoryApproved knowledge in common. A model and reviewed training for each teammate.

Support lead decides

Check the current payment record, review the draft, and send the answer from your support tool.

See the support workflow
What would you teach a teammate?

“Confirm payment settlement before promising a receipt or refund.”

Explain the correction to a teammate. Review its proposed training rule before reuse. See how teaching works →

Proposal deskRoom

Turn the discovery call into a proposal you can stand behind.

Start withDiscovery notes + approved offer + relevant past decisions

  1. ParkerProposals

    Draft the scope around the buyer’s needs.

    Claude
  2. SageTeam knowledge

    Find the approved terms and earlier decisions.

    Gemini
  3. JamieCustomer handoffs

    Give the account team a clear next step.

    OpenAI

Shared company memoryApproved knowledge in common. A model and reviewed training for each teammate.

Account lead decides

Confirm scope, pricing and commitments. Share the reviewed proposal and update the CRM.

See the sales workflow
What would you teach a teammate?

“Separate an approved offer from a customer request that still needs agreement.”

Explain the correction to a teammate. Review its proposed training rule before reuse. See how teaching works →

Client engagementRoom

Use the firm’s experience without losing the client’s context.

Start withClient brief + assigned room records + shared company methods

  1. SageTeam knowledge

    Find earlier reasoning and check its relevance.

    Gemini
  2. ParkerProposals

    Turn the brief into a proposed work plan.

    Claude
  3. MorganMarketing

    Build the client narrative around evidence.

    OpenAI

Shared company memoryApproved knowledge in common. A model and reviewed training for each teammate.

Engagement lead decides

Check what applies to this client. Review the evidence, recommendation and final deliverable.

See the consulting workflow
What would you teach a teammate?

“Explain where a past approach fits this client, and where it does not.”

Explain the correction to a teammate. Review its proposed training rule before reuse. See how teaching works →

Customer careRoom

Resolve the order question. Keep the promise realistic.

Start withOrder notes + return policy + product information

  1. RileyCustomer support

    Prepare a reply grounded in the return policy.

    Gemini
  2. JamieCustomer handoffs

    List what fulfilment needs to confirm.

    OpenAI
  3. MorganMarketing

    Draft a clearer explanation for shoppers.

    Claude

Shared company memoryApproved knowledge in common. A model and reviewed training for each teammate.

Customer care lead decides

Verify the current order and stock details. Decide on the remedy and make the update in your store tools.

See the ecommerce workflow
What would you teach a teammate?

“Check order eligibility and stock before offering a replacement.”

Explain the correction to a teammate. Review its proposed training rule before reuse. See how teaching works →

One room. Different contributions.

A customer question becomes a clear next step.

Ask one teammate to find the policy and another to prepare the handoff. You direct both contributions in the same room, with approved company knowledge behind the work.

Source connection + supplied ticket

The plan changed. The invoice did not. What happened?

  1. Bring the question

    The customer upgraded their subscription, but the latest invoice shows the old plan. They ask whether they were charged twice.

  2. Direct the two contributions

    Ask Riley to check approved billing policy. Review the facts, then ask Jamie in the same room to prepare a handoff with the missing checks and billing owner.

  3. Confirm, then send

    Billing verifies the current invoice and payment record. The rep updates the draft and sends the reply in Zendesk.

Explore support
ZendeskNotion
Them AI Prepared for review
Subscription billing handoff

The plan display does not prove a second charge. Billing will check the invoice, payment record, and any adjustment before support confirms the outcome.

Your team decides the next step

Bring a sanitized Zendesk ticket excerpt. A configured Notion connection imports policy material for approval; neither teammate looks up the live billing account.

Supplied deal notes + source connection

Parker separates the promised scope from the next request.

  1. The discovery notes

    Provide deal notes asking for a weekly catalog report and live inventory sync, alongside the approved reporting offer.

  2. The teammate’s part

    Parker uses the approved offer guidance to separate the weekly CSV reporting pilot from the live integration work that needs scoping.

  3. The deal owner’s decision

    The rep confirms what is included, asks about the missing inputs, and uses the checked outline in the buyer follow-up.

Explore sales
SalesforceConfluence
Them AI Prepared for review
Scope question

One catalog and a weekly CSV input are included. Live inventory sync needs separate discovery. Who will supply the export?

Your team decides the next step

Bring your Salesforce deal notes. Connect Confluence for source material, then approve the guidance your teammate can use.

Supplied ticket and order notes

Jamie turns an order question into a useful handoff.

  1. The order question

    Bring a delivery complaint, the order-status notes, and your approved fulfillment escalation method.

  2. The teammate’s part

    Jamie organizes the reported facts, missing shipment checks, and the next owner into a handoff draft.

  3. The operations handoff

    An authorized teammate verifies the current order and decides what to tell the customer.

Explore ecommerce
GorgiasShopify
Them AI Prepared for review
Handoff draft

Check the carrier record and fulfillment status before offering a replacement or a delivery date.

Your team decides the next step

Use ticket and order notes supplied by your team. Check current order details in Shopify before taking action.

Inside AI teammates

Set up the teammate for its part of the work.

Give it a job

Start with the work you need done.

Give Riley the policy check and Jamie the handoff. Choose each teammate’s model, instructions, room and person to run with.

Riley’s job, assigned room, model setting and person to run with

Each teammate has a job and a person to run with.01 / 03

Teach your method

A correction should help next time too.

Try a ticket example. Correct a mistaken assumption and explain why. Review the proposed training rule before that teammate uses it again.

A human correction and Riley’s draft rule awaiting review

An example, a correction, then a rule to review.02 / 03

Bring it into the room

Keep the question and the answer together.

Keep the question, cited reply and follow-up in one room. Teammates use their assigned room and approved shared company memory by default. Grant access to other rooms explicitly.

The room assignment, owner and status of a company record

Keep the room and the person responsible with the record.03 / 03

Explore the full task

Customer support

A subscription billing handoff

Sources, review and next steps Sales

Proposal preparation

Sources, review and next steps Consulting

Workshop to design brief

Sources, review and next steps

A few things worth knowing.

How do AI teammates work together?
You direct their contributions in a shared room. Ask one teammate to check the policy, review its answer, then ask another to prepare the handoff. Teammates do not automatically delegate work to each other.
Can different teammates use different models?
Yes. A teammate has its own model choice. The models available depend on your workspace configuration. Changing the model does not change its access permissions.
Is training shared with every teammate?
No. Approved training rules stay with the teammate you teach. For team-wide reuse, create a separate shared skill and review it.
Does training change the underlying model?
No. Training creates reusable instructions for that teammate. Your examples and corrections do not fine-tune the underlying model.
How is an external agent different?
An AI teammate is built and trained in Them AI. An external agent runs elsewhere and connects through MCP with its own key and retrieval scope.

Bring one real workflow

Try it on work your team already does.

Choose one recurring question, bring the source material, and name the person who will review the answer. Start there, then decide what is worth repeating.

Mac-only for now. Free during the beta (invite-only). Have a workflow in mind? Plan a pilot with us.