AgenTorQ
Experience WorkspaceWorkspace Book Demo Start free trial
Case studies · Scenarios

How a Workmate or crew works — step by step.

These are illustrative scenarios, not customer stories: generic company descriptions, the setup, each step, and the controls in play. They contain no customer names and no outcome figures.

Illustrative scenario — not a customer story

Scenarios

Six illustrative walkthroughs, by industry

Illustrative scenario — not a customer story

Illustrative scenario — not a customer storyA B2B SaaS company

Keeping renewal records current with a Sales Workmate and an Auto-Pilot crew (early access)

A mid-market SaaS sales team working in Salesforce.

AgenTorQ for SalesforceRead the walkthrough Illustrative scenario — not a customer storyA multi-region customer service organisation

Sourced case answers for a support team that works in the browser

A multi-region customer support organization using several SaaS tools and no single CRM view.

AgenTorQ Cloud (browser)Read the walkthrough Illustrative scenario — not a customer storyA global financial services firm

A governed AI rollout for a regulated operations team

An operations team in a regulated financial-services firm with a strict security review.

Both editionsRead the walkthrough Illustrative scenario — not a customer storyAn enterprise Salesforce platform team

Shipping a Salesforce change with DevSpace Studio

A small Salesforce platform team at a manufacturing company.

Both editionsRead the walkthrough Illustrative scenario — not a customer storyA leading logistics company

Answering shipment exception questions from the systems that hold the answer

A leading logistics company running a transport management system, a customer service desk and several carrier portals.

Both editionsRead the walkthrough Illustrative scenario — not a customer storyA global manufacturing company

Keeping quality and warranty records current with a Workmate and an Auto-Pilot crew (early access)

A global manufacturing company handling warranty claims and quality cases in Salesforce.

AgenTorQ for SalesforceRead the walkthrough
Walkthroughs

Challenge, setup, steps and controls

Illustrative scenario — not a customer story

A B2B SaaS company

Keeping renewal records current with a Sales Workmate and an Auto-Pilot crew (early access)

Who it describes: A mid-market SaaS sales team working in Salesforce. Edition: AgenTorQ for Salesforce

The challenge

Renewal opportunities go quiet between pipeline reviews. Next steps are stale, escalations sit in the service queue, and account managers spend the hour before each call rebuilding context from the CRM, their inbox and team chat.

The setup

  • The admin installs the AgenTorQ managed package and assigns permission sets; every read and write runs in USER_MODE.
  • Each account manager connects Microsoft 365 and Slack with their own OAuth sign-in.
  • The team describes the job in one sentence — keep renewal next steps current — and edits the Sales crew Auto-Pilot suggests (early access).
  • Guidance is written as named rules; Amount and Owner are marked never-touch; the crew works in business hours with a daily limit and is set to ask first on close dates.

How the Workmate and crew work, step by step

  1. Before a call, the account manager asks the Sales Workmate on the opportunity which renewals need attention and why.
  2. The Workmate reads the record, related cases, recent email and chat, and answers with numbered sources.
  3. It proposes a follow-up task and a new next step; both appear on an approval card with a preview of the change.
  4. On schedule, the crew checks the renewal opportunities it watches and updates next steps it is allowed to change, leaving a receipt for each.
  5. Where a close date looks wrong, the crew asks first; the account manager rejects one suggestion, and the crew keeps that as a lesson.
  6. The team lead reviews the live feed and the job board, where a failed job shows its reason and a Try again button.

The controls in play

ApprovalsRecord changes from the Workmate wait for the account manager.
MaskingNamed personal data is masked before any prompt leaves.
Audit trailEvery prompt, read, approval and write is recorded.
Receipts and undoEach crew change leaves a receipt and can be undone.
Illustrative scenario — not a customer story

A multi-region customer service organisation

Sourced case answers for a support team that works in the browser

Who it describes: A multi-region customer support organization using several SaaS tools and no single CRM view. Edition: AgenTorQ Cloud (browser)

The challenge

Answers live in knowledge articles, shared drives and past tickets. Agents search each tool separately and replies vary from person to person.

The setup

  • Agents sign in to AgenTorQ Cloud in the browser — nothing to install.
  • An admin builds a team WorkSpace with profiles from the feature-toggle grid and shares the Service Workmate and the knowledge base.
  • Help articles and policy documents are added to the RAG knowledge base as folders, files and chunks; personal data is excluded or masked before it is embedded.
  • Each agent connects the ticketing and chat tools they already use, with their own sign-in.

How the Workmate works, step by step

  1. An agent asks the Service Workmate how to handle a customer's configuration question.
  2. The Workmate searches the knowledge base and connected tools, then answers with citations to the exact passages.
  3. It drafts a reply; the agent edits the wording before sending.
  4. The agent asks it to update the ticket; the change is shown on an approval card and runs only after approval.
  5. The support lead reviews which sources shaped answers and refines the knowledge base.

The controls in play

Cited answersAnswers come from approved sources scoped to the agent's access.
ProfilesDeny-by-default access, with per-user overrides.
ApprovalsTicket updates and sends wait for the agent.
IsolationPersonal chats and tokens stay private to each agent.
Illustrative scenario — not a customer story

A global financial services firm

A governed AI rollout for a regulated operations team

Who it describes: An operations team in a regulated financial-services firm with a strict security review. Edition: Both editions

The challenge

Staff want AI help with routine operations work, but security will not approve tools that send customer data to a model without masking, policy or a record of what happened.

The setup

  • Security reviews the deployment options, including a dedicated, network-isolated deployment and a self-hosted connector runtime.
  • Admins create roles from the feature-toggle grid, deny-by-default, and set model allowlists per role and workflow.
  • Provider keys are added under the firm's own accounts (bring your own key), with an OpenAI-compatible or self-hosted endpoint for sensitive work.
  • Database connections are added through the credential vault with read-first, parameterized SQL.

How the Workmate works, step by step

  1. An analyst asks the Ops Workmate to reconcile a list against the operational database.
  2. The request is checked against role, model and data policy before anything runs; named personal data is masked.
  3. The Workmate answers with results cited to the rows it read.
  4. It proposes a correction; the DML guard holds the write until an approver accepts it.
  5. Security exports the audit trail via API into its existing tooling for review.

The controls in play

RBACAccess is server-enforced; the client can request, only the server decides.
Model allowlistsOnly approved models are callable for this workflow.
DML guardDatabase writes are held until approved.
Exportable auditActor, role, route, data touched and approval, via API.
Illustrative scenario — not a customer story

An enterprise Salesforce platform team

Shipping a Salesforce change with DevSpace Studio

Who it describes: A small Salesforce platform team at a manufacturing company. Edition: Both editions

The challenge

Small Apex and Flow changes queue up behind larger work because each one means switching between the editor, the org, git and the test runner.

The setup

  • Developers open DevSpace Studio and connect the org and the GitHub repository.
  • The team lead sets the operating mode per task: Plan for exploration, Edit for changes; Plan never deploys and deletes always confirm.
  • The Admin Workmate is available for questions about schema and validation rules, grounded in org metadata.

How the Workmate works, step by step

  1. A developer asks the Developer Workmate to add a risk score to renewal opportunities.
  2. The Workmate reads the code and proposes a surgical diff to the Apex class and its test.
  3. It runs the Apex tests and shows the results next to the diff.
  4. The developer approves; the Workmate deploys and opens a pull request on GitHub.
  5. If a test fails after deploy, the Workmate proposes a fix for review in the same window.

The controls in play

Operating modesAsk, Edit, Plan, Auto and Bypass, enforced per turn.
Approve before deployPreview-then-apply on every write and deploy.
Real gitBranches, commits and pull requests on GitHub.
Audit trailEvery tool call and deploy is recorded.
Illustrative scenario — not a customer story

A leading logistics company

Answering shipment exception questions from the systems that hold the answer

Who it describes: A leading logistics company running a transport management system, a customer service desk and several carrier portals. Edition: Both editions

The challenge

When a shipment goes wrong, the answer is spread across the transport system, the carrier update, the email thread and the service ticket. Co-ordinators rebuild the story by hand for every call, and different people give different answers.

The setup

  • Co-ordinators sign in to AgenTorQ Cloud in the browser and connect the systems they already use, each under their own sign-in.
  • Operational databases are connected through the credential vault with read-first, parameterized SQL and a DML guard on writes.
  • An admin builds a team WorkSpace, shares an Operations Workmate, and sets model allowlists for the role.
  • Standard operating procedures are added to the knowledge base; personal data is excluded or masked before anything is embedded.

How the Workmate works, step by step

  1. A co-ordinator asks the Operations Workmate what happened to a delayed shipment and what was promised to the customer.
  2. The Workmate reads the shipment record, the linked ticket, the carrier message and the procedure, under that co-ordinator’s own access.
  3. It answers with numbered sources, so each statement can be traced to the record or message it came from.
  4. It proposes a ticket update and a customer note; both wait on an approval card with a preview of the change.
  5. After approval the update is written back, and every step of the request is recorded in the audit trail.

The controls in play

Read-first SQLParameterized, read-first queries with a DML guard on writes.
MaskingNamed personal data is masked before any prompt leaves.
ApprovalsTicket and record updates wait for the co-ordinator.
Audit trailActor, sources, approval and write, on the record.
Illustrative scenario — not a customer story

A global manufacturing company

Keeping quality and warranty records current with a Workmate and an Auto-Pilot crew (early access)

Who it describes: A global manufacturing company handling warranty claims and quality cases in Salesforce. Edition: AgenTorQ for Salesforce

The challenge

Quality cases arrive from dealers, service partners and the plant. Case records go stale between reviews, and engineers spend the start of every meeting working out which claims relate to the same part.

The setup

  • An admin installs the managed package and assigns permission sets; every read and write runs in USER_MODE.
  • Engineers connect the document store and email they already use, each with their own sign-in.
  • The team describes the job in one sentence and edits the crew Auto-Pilot suggests (early access).
  • Guidance is written as named rules: warranty amounts and account owners are never-touch, the crew works in plant hours with a daily limit, and it asks first before changing a case status.

How the Workmate and crew work, step by step

  1. An engineer asks the Workmate on a quality case which open claims involve the same part and what has already been tried.
  2. The Workmate reads the related cases, the attached reports and the email thread, and answers with numbered sources.
  3. It proposes a summary on the case and a follow-up task; both appear on an approval card with a preview.
  4. On schedule, the crew refreshes the fields it is allowed to change on the cases it watches, leaving a receipt for each.
  5. Where a case status looks wrong, the crew asks first; a rejected suggestion becomes a lesson it keeps.
  6. The quality lead reviews the live feed and the job board, where a failed job shows its reason and a Try again button.

The controls in play

USER_MODECRUD, field-level security and sharing are always enforced.
Never-touch fieldsNamed fields are out of scope for the crew.
Receipts and undoEach crew change is reversible from its receipt.
Working hours and limitsDaily limits, plant hours and auto-pause.
Customer stories

Customer stories will appear here.

We publish a customer story only when the customer has agreed to it. Until then, this page shows illustrative scenarios.

Share your story, or become a design partner.

Tell us how you would use AgenTorQ and Auto-Pilot crews (early access). We will reply from reachout@agentorq.com.

Become a design partner Share your story