Noice Assistant
An AI assistant in Microsoft Teams that monitors your sites, drafts the work and reports back. Publishing stays with your team.
- CVE2 critical advisories ticketed to JiraDone
- SEOWeekly audit posted. 3 fixes recommendedDone
- WCAGAccessibility scan clean on 12 changed pagesDone
- CMS2 page drafts awaiting your approvalYours
- REPORTMonthly status report delivered to SharePointDone
Every action logged. Publishing stays with your team.
Not another chatbot
Chatbots wait to be asked. Noice Assistant turns up with the work already done.
Typical AI chatbot
- Answers questions when someone remembers to ask
- No access to your service desk, CMS or analytics
- No idea what changed on your sites overnight
- Trusted with nothing, so it does nothing
- Shared platform, your data in someone else’s pool
Noice Assistant
- Watches your sites, stack and tickets on a schedule
- Works your actual tools: Jira, your CMS, SharePoint, analytics
- Posts digests, audits and alerts before you ask
- Drafts the work; publishing stays with your team
- Dedicated single-tenant deployment
Scoped, deployed, operating
01 / Scope
We map your estate
Your sites, CMS, service desk, and the stack behind them go into an asset register, along with what the assistant is allowed to do and for whom.
02 / Deploy
Your assistant, your tenant
A dedicated, single-tenant deployment, wired into your Microsoft Teams. Nothing shared with other clients.
03 / Operate
It runs the routine
Daily digests, scheduled audits, security alerts and monthly reports arrive on their own. Chat with it in Teams for everything in between. We maintain and improve it monthly.
One assistant, six jobs
Every module is the same assistant wearing a different hat: scheduled, scoped, and reporting back in Teams. Run the ones that matter to you.
Service desk
Raise, triage, comment and transition Jira tickets from chat. Anything outside scope is logged as a ticket for a human, never dropped. Answers questions from your own docs and procedures.
SEO watch
A weekly technical and content audit of your sites: crawl health, schema, performance, rankings and Search Console data, delivered as a branded report to Teams with fixes ranked by impact.
Accessibility watch
WCAG 2.2 audits across your pages with findings ranked by user impact. Know your conformance position before an audit or a complaint finds it for you.
Security watch
Daily critical-CVE monitoring against your actual stack and a monthly end-of-life review. New advisories are ticketed to Jira automatically, with no duplicates and no silent days.
Content operations
Drafts pages and edits in your CMS from a chat message or an attached brief, and hands you a preview link. Publishing is always a human decision.
Reporting
A client-ready monthly status report, generated from your live delivery data, on brand, delivered to SharePoint and summarised in Teams. Quarterly review packs on request.
The first demo runs on your public website. No access required.
Start with a two-week sprint
A discovery workshop maps where AI can help across your teams. We then build the agent the workshop chooses and embed it in Microsoft Teams or Slack, wired into the tools you already use. You finish the fortnight with:
- One live agent doing a real job your team does today
- An opportunity map across your teams, scored by value and effort
- A plain-English runbook for the people who supervise it
- A path into a managed subscription, without re-scoping
Built for organisations that have to say no to most AI tools
The interesting question is not what an AI assistant can do. It is what yours provably cannot. Security posture documentation is available for your IT review.
Publishing stays with your team
The assistant cannot publish content, and it can only email addresses you approve.
Full audit trail
Every tool call is logged: what was read, what was created, what was blocked. A monthly audit report of everything your assistant did is available as an add-on.
Single-tenant deployment
Each client gets a dedicated deployment. Model calls go through Amazon Bedrock in Australia, and your data is not used to train models.
Your choice of model
The assistant runs on the AI model your organisation approves.
Frequently asked questions
01
What is Noice Assistant?
Noice Assistant is an AI operations assistant that lives in your Microsoft Teams and runs the routine work of a digital estate: service desk triage, SEO and accessibility monitoring, security watch, CMS content drafting and monthly reporting. It is deployed as a dedicated, single-tenant service per organisation and maintained by Noice as a managed subscription.
02
Where does it run and where does our data go?
Each client gets a dedicated deployment. Model calls go through Amazon Bedrock in Australia, and your data is not used to train models.
03
Can it take destructive actions in our systems?
The assistant cannot publish content, and it can only email addresses you approve.
04
How do we know what it has been doing?
Every tool call is written to an audit log: what was read, what was created, what was refused. Scheduled jobs are never silent; if a daily security check finds nothing, it says so. A monthly audit report of all assistant activity is available as an add-on.
05
What does it integrate with?
Microsoft Teams for chat and delivery, Jira and Confluence for service desk and documentation, SharePoint for reports, your CMS or DXP for content operations (Magnolia, Adobe Experience Manager and others), and Google Search Console and Analytics for SEO. If your estate runs on something else, integrations are scoped per client.
06
Do we have to take every module?
No. Modules are bolt-on. Your subscription is scoped to the modules you run.
07
How long does deployment take?
A scoped demonstration against your public website can be ready in days, because the SEO, accessibility and security audits need no access to your systems. Production timing depends on your IT security review.
08
Who controls what our staff can ask it to do?
Access is allowlisted: only nominated people in your organisation can use the assistant, and its scope of work is agreed and enforced per deployment. Requests outside that scope are politely declined and logged as tickets for a human.
09
Is this a chatbot?
It chats, but the value is in what it does unattended: scheduled audits, security monitoring, ticket hygiene and reporting that arrive without anyone asking. The chat surface is how you steer it and ask for more.