Agentic AI for secure behaviour management
Your secure behaviour management programme, run by AI agents you control.
Agents run multi-channel phishing simulations, training, nudges, reporting and phishing response, in your policy and tone. You decide what needs approval.
From one request to a ready simulation
The agent asks, plans and builds. You approve in the middle.
From the console A recording in Keepnet’s demo tenant, with example data. The agent asked three questions and built nothing until the plan was approved; then it uploaded the simulation and assigned it to a group, where it waits for approval.
Security teams running Keepnet
Programme work moves from your list to the agents
Agents read each employee’s behaviour and history, then explain each choice before you approve it.
The agent drafts from your own documents, never names the employee, and waits while you edit the plan.
Recurring phishing, QR code and training activities run on schedule, at a sensible hour in each employee’s time zone.
When you switch them on, agents analyse what employees report and act on intent, under your playbooks.
Ask in plain English. The agent picks the view that fits the question: KPI dashboard, line, bar, heatmap, Sankey and more.
Agents read each employee’s behaviour and history, then explain each choice before you approve it.
One assistant behind these outcomes
- Built-inspecialist agents, picked for each request
- Your ownagents with the instructions, tools and documents you choose
- Every planincluded, with fair use
Screens from Keepnet’s demo tenant, 12 October 2026.
More on reports from the Reporting Agent and reported email triage.
Content in your policy, your tone and your context
Agents read the documents and organisation details you give them, and show which ones they used.
Your documents
Policies, procedures and guidelines in PDF, Word, Excel, Markdown or text, up to 50 MB each. Upload a folder or a ZIP.
A knowledge base you can check
Ask a question and see exactly which passages an agent retrieves, with the source document.
Organisation memory
Sector, employee count, languages, brand tone and priority risks, shared across your team.
A specialist for each job
The Main Assistant sends each request to the right specialist. You can also pick one yourself.
Swipe to see more
Microlearning Agent
Creates short training modules for compliance and security education, with translations.
Phishing Email Generator
Creates phishing email simulations and landing pages from behaviour profiles.
Smishing SMS Assistant
Creates SMS phishing simulations and their landing pages.
Vishing Call Assistant
Places two-way voice phishing simulation calls with an AI voice.
Deepfake Video Assistant
Creates deepfake video simulations for training. Opt-in only.
Voice Clone
Creates voice clones for your catalogue, only with the speaker’s consent.
Recommendation Assistant
Recommends existing library items for each employee.
Policy Summary Assistant
Answers questions on your security policies and summarises them.
Report Agent
Creates branded reports from your platform data.
SBM Analyst
Read-only analyst for secure behaviour data across employees and devices.
Incident Responder Assistant
Analyses the reported email queue.
Scheduled Tasks Assistant
Sets up and runs recurring phishing and training tasks.
1 of 12
An agent for your own process, tested before anyone uses it
Describe it, keep only the tools and documents it needs, and test it. It runs in chat only once you switch it on.
Name the agent and say what it should do.
Draft with AI turns the description into instructions you can edit.
Draft with AI suggests tools. Keep what the job needs; each is marked read-only or elevated access.
Point it at the policies it should answer from, or leave it on the whole knowledge base.
Send a sample message and read the reply. Nothing is saved until you create the agent.
Name the agent and say what it should do.
From the console The Create agent form in Keepnet’s demo tenant (ACME Manufacturing). The agent was tested, not saved. A Builder mode also drafts the agent from a short conversation.
Approval by default. Autonomy only where you allow it.
Safeguards apply in both modes and cannot be switched off.
Approval-gated execution
The default. Agents recommend, and actions run once an admin approves. Approve a whole run or one employee at a time.
Autonomous execution
Your choice, when you are ready. Actions run on their own within the policies you set.
Operational safeguards
Simulation frequency cap, training cooldown window, holiday and blackout periods, and each employee’s time zone.
Ethics and governance
Ethical AI communication, alignment with your governance, inclusive language, equality and non-discrimination.
Clear and culturally aware
Plain language, accessibility-aware wording, transparent AI communication and culturally sensitive messages.
Deepfake and voice
Deepfake is opt-in. A voice is cloned only with the consent of the person it belongs to.
From the console Company Settings, Agentic AI Settings. Read how the settings work
Questions to ask before agents run your programme
Based on the AI control questions Gartner advises CISOs to put to vendors.7
| Question | Keepnet |
|---|---|
| Does anything run without a person approving it? | Not unless you allow it. Approval is the default, and autonomous execution runs inside mandatory safeguards. |
| Can we start with a pilot group? | Yes. Scheduled tasks and campaigns target the groups you choose. |
| What stops it sending too much? | A simulation frequency cap, a training cooldown window, holiday and blackout periods, and each employee’s time zone. |
| How is the content kept accurate and on-policy? | It is grounded in your own documents, the chat lists the sources it used, and you edit the plan before anything is generated. |
| Can it create a deepfake of our executives? | Only if you opt in, and only with the consent of the person whose voice or face is used. |
| Is our data used to train the model? | No. Customer data is never used to train or improve any model, Keepnet’s or a provider’s. |
| Does personal data reach the model? | No. Personal data stays in your tenant and is removed before AI processing. |
Your data stays yours
And the model key can be yours too.
No personal data to the model
Personal data stays in your tenant and is removed before AI processing.
Never used for training
Not by Keepnet and not by any provider. It is in the contract, in the data processing agreement.
Hosted in Microsoft Azure
GPT models run in Microsoft Azure OpenAI Service, and nothing is sent to OpenAI. Your data stays in the region you choose.
Your own key, if you prefer
Bring your own provider key. It is stored in Cloudflare Secrets Store and never shown again after saving.
Costs in view
Each answer shows its tokens and the cached share. Usage and cost breaks down by user, model and product, with CSV export.
ISO/IEC 42001 certified
Keepnet’s AI management system is certified to ISO/IEC 42001:2023, alongside ISO 27001, 27017 and 27018.
"It would be unfair to single out just one feature. The mail scenarios and customisable training modules are particularly excellent."
"It's a really great platform with a lot of flexibility to ensure our requirements are met. Onboarding was friendly, enjoyable, and thorough."
"It's super easy and intuitive to setup campaigns, whilst making it easier for end-users to report phishing emails."
"The product itself works great on day one and delivers immediate value as our team was able to get started with our phishing exercise program immediately."
"It has been consistent, reliable, and implementations over time shows that they really listen customer feedback."
"We've seen a noticeable improvement in how employees react to suspicious emails or phone calls- because they remember the stories."
"Within minutes, we can create realistic, scenario-based templates specific to our organization, reducing setup time while increasing the relevance of training."
"The simulations and campaigns are very customisable and realistic."
"Thanks to Keepnet's Incident Response tool, we reduced the e-mail analysis and response time, which used to take hours, to 2 minutes."
Gartner's notice on these reviews is at the foot of the page.
Keepnet Labs has a rating of 4.9 out of 5 based on 104 ratings in the Security Awareness Computer-Based Training market on Gartner Peer Insights as of October 2026. Gartner and Peer Insights™ are trademarks of Gartner, Inc. and/or its affiliates. All rights reserved. Gartner Peer Insights content consists of the opinions of individual end users based on their own experiences, and should not be construed as statements of fact, nor do they represent the views of Gartner or its affiliates. Gartner does not endorse any vendor, product or service depicted in this content nor makes any warranties, expressed or implied, with respect to this content, about its accuracy or completeness, including any warranties of merchantability or fitness for a particular purpose.
Recent additions
From the 2026 release notes.
- AI Ally picks a scenario for each user much faster, so large campaigns start on time.
- Agentic AI microlearning editing can swap videos, remove pages and change questions, and vishing calls can run without recording.
- Agentic AI reports follow the platform palette and keep chart text in PDF export.
- Deepfake video in AI microlearning, and Scheduled Tasks for recurring phishing and training.
Agentic AI pricing
Included in every plan
With fair use and no per-task charge. Bring your own model key if you prefer.
What buyers ask about Agentic AI
Short answers, with the detail below each one.
AI agents that plan, create, deliver and measure the work of your programme, rather than a chatbot that only answers questions.
In Keepnet they run your secure behaviour management programme: multi-channel phishing simulations, training, nudges, reporting, and the analysis of reported emails, SMS and calls. You decide what needs your approval.
Only if you allow it. Approval-gated execution is the default.
Autonomous execution runs inside mandatory safeguards, such as a simulation frequency cap and blackout periods.
No. Keepnet does not use customer data to train, fine-tune or improve any model, its own or a third party’s, and its AI sub-processors are contractually prohibited from doing so.
GPT models hosted in Microsoft Azure OpenAI Service, and open-weight models on Cloudflare Workers AI. Nothing is sent to OpenAI.
You can also bring your own provider key.
Yes. Admins create company agents with their own instructions, model, tools and documents, test them, then make them available in the chat.
Yes. Upload your policies and guidelines, switch them into the knowledge base, and agents use them to answer questions and to create training and simulations. The chat lists the sources it used.
No. It takes routine programme work off the team, and the team keeps the decisions it wants to keep.
It is included in every Keepnet plan, with fair use. There is no per-task charge.
See Keepnet with your organisation in mind
- See the agents on your own programmeTraining, simulations and reports, in your policy and tone.
- Follow one request end to endFrom the questions to the plan to the preview.
- Set the controlsExecution mode, safeguards and your own model key.
- Connect your own toolsIf it fits, a proof of value follows, on your timeline.
Book a demo
Thirty minutes, no cost and no obligation.
Thank you, your request is in
A confirmation is on its way to your inbox.
Thirty minutes with the Keepnet team, on Microsoft Teams. Your details are already filled in.
Prefer to pick a time later? The team will follow up within one working day.
Your demo is booked
A calendar invite is on its way to your inbox, with a Microsoft Teams link.
The booking calendar will not open in this page.
Open the calendar in a new tabSources
- Keepnet console, dash.keepnetlabs.com, Keepnet demo tenant, read 12 October 2026: Agentic AI chat, Agents, Create agent, Files, Memory, Usage and cost, Activities, Scheduled Tasks and Agentic AI Settings.
- Agentic AI, doc.keepnetlabs.com, October 2026. doc.keepnetlabs.com/next-generation-product/platform/agentic-ai
- Agentic AI Settings, doc.keepnetlabs.com, October 2026. doc.keepnetlabs.com/.../agentic-ai-settings
- AI Transparency, Keepnet legal hub, October 2026. doc.keepnetlabs.com/legal-hub/for-customers/ai-transparency
- Data Processing Agreement, section 12, Keepnet legal hub. doc.keepnetlabs.com/legal-hub/for-customers/data-processing-agreement
- Release notes 2026, doc.keepnetlabs.com. doc.keepnetlabs.com/resources/release-notes/2026
- Gartner, “Ditch Security Awareness and Adopt AI-Powered Secure Behavior Management”, William Candrick, Alex Michaels, G00860839, 22 September 2026. GARTNER is a registered trademark of Gartner, Inc. and/or its affiliates and is used herein with permission. All rights reserved.