tenmoregate™ Start a free trial
Use cases

How teams put Tenmore Gate to work.

Five situations most organizations meet as AI spreads, and how they play out with Tenmore Gate, from the person responsible to the result.

01 · MODEL ROLLOUTSRoll out a new model without surprises 02 · INTERNAL AGENTSAn agent for every team 03 · PUBLIC AGENTSAnswer your website’s visitors, within limits you set 04 · CODING AGENTSGive developers coding agents, within limits 05 · SENSITIVE WORKKeep sensitive work on your own hardware
Use case 1 · Model rollouts

Roll out a new model without surprises

Head of IT, 600-person insurerAs the IT lead, I want people to get a new model only after it has been tested on our own work, so that a model change never surprises 600 people at once.

Before

A provider releases a better model and every team asks for it. Switching everyone at once is a gamble, and going back means another round of changes.

With Tenmore Gate

  1. Add the new model to the Support gateway as a draft. Nobody’s work changes yet.
  2. Run the support test suite on staging: 38 of 40 answers pass, against 34 for the current model.
  3. Release it to a pilot group of 20 people and compare failures, speed and cost with production for a week.
  4. Promote it to everyone, or roll back in one step. Every step is in the audit log.

Result. The new model reached everyone after a week of evidence, not a launch-day guess.

Draftnew model addednobody affected Stagingadmins only38 of 40 tests pass Pilot group20 people, a weekvs. production Everyonepromotedin the audit log roll back in one step, at any stage
From draft to everyone, with a way back at every stage.
Stages of the Support gateway: the new model in staging and in rollout to a pilot group, compared with production
The new model in staging and with the pilot group, compared with production.
Runs of the support test suite on staging and production
The same test suite on staging and production before promoting.
Use case 2 · Internal agents

An agent for every team

IT for HR, finance and the IT helpdeskAs an IT admin, I want to give teams agents that know their own documents, so that people stop pasting handbooks into chatbots on personal accounts.

Before

HR answers the same questions every week. People who want a quicker answer build their own assistants on personal accounts and upload company documents to them.

With Tenmore Gate

  1. Upload HR’s handbook. Tenmore Gate turns it into skills (leave, expenses, benefits) and tests the agent against the handbook.
  2. Share the HR assistant with everyone, and the finance agent with the finance team only.
  3. People open the agent’s chat link and ask. Answers come from the handbook, on approved models, with the same filters as everything else.
  4. When the handbook changes, upload it again. Tenmore Gate rebuilds the skills and tests them, showing the new score next to the last one before IT updates the agent.

Result. Each team has an agent that knows its documents, and IT still decides what runs where and who uses it.

Handbook.pdfHR Expenses policy.docxFinance IT runbook.pdfIT Skills leave and holidaysexpense claims benefits · laptopstested on the documents HR assistanteveryone in the company Finance agentfinance team only IT helpdeskeveryone in the company
Documents become skills; each agent is shared with its own audience.
The HR assistant in Chat: a greeting and three suggested questions
What people see: the HR assistant in Chat.
The Agent section of the HR assistant gateway: instructions, three skills and the chat page address
What IT sees: instructions, skills at fixed versions, and the chat link.
The skills library with skills for HR and the website
The skills library: create from documents, write by hand, import or download.
Use case 3 · Public agents

Answer your website’s visitors, within limits you set

Marketing lead, software companyAs the marketing lead, I want visitors to get answers on our website at any hour, so that their questions turn into trials, without surprises in cost or in what the agent says.

Before

Visitors leave with unanswered questions outside office hours. A chatbot vendor would mean another contract, another data processor and no control over the model.

With Tenmore Gate

  1. Build the agent from public documents only: product sheets, prices and the FAQ.
  2. Turn on public access for the company website, with a monthly budget that stops it and limits per visitor.
  3. Add one line to the website. Every chat says it’s an AI, and the agent can’t reach company tools.
  4. Watch questions and cost in Monitoring, and switch it off at once if needed.

Result. Visitors get answers day and night, on your models and your terms. The assistant on tenmore.com runs exactly this way.

Visitorno account Your websiteone line of code Tenmore Gate A budget that stops it Limits per visitor AI disclosure, always Public skills only Modelyour account
Every visitor’s message passes the same checks before the model answers.
The Tenmore Gate product page with the Ask about Tenmore Gate assistant open beside it
The assistant on tenmore.com, a public agent on Tenmore Gate.
Public access settings of the Website assistant: limits per visitor, the address and the embed code
Public access: limits, the address and the line to add to your website.
Use case 4 · Coding agents

Give developers coding agents, within limits

Engineering manager, fintechAs the engineering manager, I want developers to use AI coding agents on our repositories, so that they move faster without an agent merging to main or leaking secrets.

Before

Developers already use coding agents with their own accounts and tokens. Nobody knows what those agents can change, or what they cost.

With Tenmore Gate

  1. Developers point their coding tools at the Engineering gateway, each with a personal key.
  2. Write tool rules in plain English, such as “No merges to main from agents”. Secrets are redacted from tool results before the model reads them.
  3. See each developer’s cost, and cap the month with a budget.
  4. When someone leaves, their key stops working at once.

Result. Agents do the routine work; the rules decide what they can’t touch, and every call is on record.

  • ✓ github__search_code(query: "retry policy")14 results · run as sam@acme.com
  • ✓ github__get_file(path: "config/payments.yml")result redacted: 1 × [secret]
  • ✕ github__merge_pull_request(repo: "payments", number: 912)Blocked by policy: "No merges to main from agents"
  • ✓ jira__add_comment(issue: "PAY-318")allowed · logged
One agent session: tools run as the developer, a secret redacted, a merge to main blocked.
Monitoring: calls, failures, policy blocks and cost per gateway for a week
Calls, policy blocks and cost per gateway and person.
Use case 5 · Sensitive work

Keep sensitive work on your own hardware

Legal lead, manufacturerAs the legal lead, I want to use AI on contracts without them leaving our network, so that confidentiality doesn’t depend on trusting a provider.

Before

Contracts can’t go to an outside provider, so the legal team does without AI while everyone else uses it.

With Tenmore Gate

  1. Tenmore Gate runs inside the company network, next to a model on the company’s own server.
  2. The Legal gateway is open to two people and reaches only that model.
  3. Redaction removes names, account numbers and other personal data before any model sees them, and a policy blocks customer financial details.
  4. Other teams keep using cloud models for everyday work, under the same rules.

Result. Legal uses AI on contracts, and the contracts never leave the building.

Your network Legal team2 people Tenmore Gate redacts names andaccount numbersblocks financial details GPU servercontracts model Other teamseveryday work Cloud modelsother teams only contracts never leave
Inside the network: Legal reaches only the company’s own model.
Five team gateways; Legal is open to two people and uses a model on the company’s own servers
Legal’s gateway: two people, one on-premises model.
A gateway’s filters: a plain-English policy, redaction of personal data and an output filter
Redaction and a plain-English policy, in the order a request meets them.

Which of these is yours?

Try Tenmore Gate on your own models and teams, or ask us how it would fit.

Start a free trial Ask a question