The Week in Workplace Tech: AI Agents Move Into the Team, While Identity Risk Grows

Workplace technology is becoming more capable, more connected and more difficult to govern. This week’s news shows why business leaders need to treat productivity, collaboration and cybersecurity as one conversation rather than separate technology projects.

Introduction

From Australia to the wider global market, organisations are testing AI agents that can work across projects, documents, conversations and business systems. The promise is significant: less administration, faster decisions and more consistent customer service.

The risk is growing at the same time. AI agents need access to data and applications, while attackers are increasingly targeting trusted identities, collaboration platforms and human behaviour. For small and mid-sized businesses, the practical lesson is simple: productivity gains must be matched with clear permissions, visible accountability and reliable recovery plans.

This week’s main developments point to five priorities for managers and IT decision makers: govern AI before it spreads, secure collaboration tools, strengthen identity controls, review software supply chain exposure and measure technology by business outcomes rather than novelty.

1. AI agents are moving from personal assistants to team participants

Atlassian used its Team 26 Europe event on October 6 and 7 to introduce the Agentic Multiplayer Protocol, or AMP. The company describes it as a framework for collaboration between people and AI agents across its platform. The announcement is part of a wider shift in workplace software, where AI is expected to work inside shared projects and workflows rather than remain confined to an individual chat window.

Atlassian says AMP is designed to connect human and agent activity across tools such as Jira, Confluence and external applications. Agents may be able to create work, update records, summarise progress and pass tasks between systems, with activity recorded in shared workspaces. Atlassian’s announcement provides the vendor’s view of the change, while industry coverage from Computer Weekly highlights the wider movement towards agentic work management.

Why it matters for businesses

For a growing business, an agent that can find information, update a project board and prepare a customer response could remove hours of low-value administration. It could also reduce delays caused by information being scattered across email, chat, spreadsheets and line-of-business systems.

However, an agent with write access is not just another software feature. It is a new type of business identity. If it is poorly configured, it may expose confidential information, make an unauthorised change or act on outdated instructions. The risk becomes more serious when multiple agents interact without a clear owner or approval process.

Practical recommendations

• Start with one or two low-risk use cases, such as meeting summaries, internal knowledge search or draft project updates.
• Create an AI register that records which tools are approved, what data they can access and who owns each deployment.
• Give agents the minimum permissions required and separate read access from write or approval rights.
• Require human approval for payments, contract changes, customer commitments, security settings and other high-impact actions.
• Keep an audit trail showing what an agent did, which data it used and which person approved the result.

The important question is not whether an AI agent is clever. It is whether the business can explain, limit and reverse its actions.

2. Collaboration platforms are now part of the attack surface

Microsoft has warned about a campaign in which threat actors impersonate IT support staff through Microsoft Teams. The attackers use external collaboration features to contact employees, persuade them to approve remote access and then use legitimate tools such as Quick Assist or remote management software to gain control of a device.

Microsoft’s analysis says the activity is not a technical compromise of Teams itself. Instead, it exploits trust, urgency and the appearance of normal support activity. Once access is granted, the attacker may use PowerShell, install malicious packages, collect information about the environment and attempt lateral movement. The Microsoft Security advisory includes detection and mitigation guidance.

Why it matters for businesses

Many organisations have invested heavily in email filtering but have not applied the same level of discipline to chat, video meetings and collaboration tools. Staff may be more willing to trust a message that appears inside Teams, Slack or another familiar platform, particularly when it appears to come from someone offering urgent technical assistance.

For small businesses without a large internal IT team, this type of attack is especially concerning. Employees may be accustomed to receiving support from external providers, contractors or managed service partners. That makes it essential to verify the person and the process, not simply the name shown in a chat window.

Practical recommendations

• Tell staff that legitimate IT support will never ask them to bypass security warnings or share passwords and one-time codes.
• Use a known support channel and published phone number to verify unexpected requests for remote access.
• Review external messaging settings in Teams and other collaboration platforms, including guest access and cross-tenant communication.
• Require managed devices, strong MFA and device compliance checks before sensitive remote support sessions.
• Monitor for remote support tools followed by PowerShell, command-line activity, script execution or unusual network connections.

The most effective control is a simple one: make the safe process easier and more familiar than the attacker’s process.

3. Cisco and other vendors are building security into agentic collaboration

Cisco announced new agentic collaboration capabilities at WebexOne on October 7. The company described tools intended to connect conversations, enterprise context and business actions while keeping identity, security, governance and visibility in the same control framework. Cisco’s announcement reflects a broader vendor effort to make AI collaboration useful without leaving governance as an afterthought.

The direction is significant because business users rarely work inside one application. A customer request may begin in a meeting, move into a task board, require a document update and finish with an entry in a customer relationship system. AI agents are being designed to move through that chain on behalf of employees.

Why it matters for businesses

Integrated workflows can reduce duplication and improve response times, but they also create concentration risk. A single compromised account or badly configured agent may provide a path across several systems. The business may not know whether an action was taken by a person, an automated workflow or an external integration.

Vendors will continue to promote built-in governance, but businesses remain responsible for deciding what is appropriate for their data, customers and regulatory obligations. This is particularly relevant for Australian organisations handling personal information, health data, financial records or sensitive commercial information.

Practical recommendations

• Map where sensitive information moves between email, chat, meetings, storage, CRM and finance systems.
• Ask vendors how agents are authenticated, how permissions are inherited and how actions are logged.
• Use separate service identities for integrations instead of sharing a human administrator account.
• Apply data loss prevention and retention policies to collaboration content, recordings and AI-generated summaries.
• Test what happens when an agent is disabled, a user leaves the business or an integration is compromised.

A secure collaboration strategy should make it clear who can see information, who can change it and who is accountable when automation makes a mistake.

4. Microsoft’s 2026 threat report puts identity and interconnected risk at the centre

Microsoft’s 2026 Digital Defense Report, published on October 1, describes a threat environment shaped by connected cloud services, software supply chains, digital identities and AI. The report says attackers are using AI to increase speed and scale, while people remain a heavily exploited path into organisations through phishing, impersonation and social engineering.

The report also emphasises phishing-resistant MFA, passkeys, disciplined identity management, tiered administration and strong controls for privileged access. Microsoft identifies open-source supply chain compromise, attacks against edge devices and AI-assisted malicious activity as major areas to watch. The full report provides the detailed findings and recommended priorities.

Why it matters for businesses

Small and mid-sized organisations often rely on a mixture of cloud applications, contractors, outsourced IT and older devices. That connected environment can be efficient, but it can also make it difficult to see which accounts, applications and suppliers have access to important systems.

The report’s most practical message is that identity is now the control plane for security. Firewalls and antivirus remain important, but they cannot compensate for excessive permissions, unmanaged accounts or weak authentication. When an employee, supplier or application identity is compromised, attackers may be able to move through trusted services without triggering obvious alarms.

Practical recommendations

• Enforce MFA across email, remote access, administration and finance systems, prioritising phishing-resistant methods where available.
• Remove dormant accounts, shared logins and unnecessary administrator privileges.
• Review external users, contractors, application permissions and OAuth connections at least quarterly.
• Separate administrator accounts from everyday email and web browsing accounts.
• Back up critical data in a way that prevents ordinary user accounts from deleting or encrypting the backups.

Businesses do not need a perfect security architecture to reduce risk. They do need a reliable process for knowing who has access and removing access when it is no longer justified.

5. The software supply chain deserves more attention from smaller organisations

Supply chain risk is no longer limited to large technology companies. Modern businesses depend on cloud platforms, browser extensions, open-source components, managed service providers, payment systems and software integrations. A weakness in one of those relationships can create consequences for many customers at once.

Microsoft’s latest threat reporting places open-source supply chain compromise among the most important emerging risks. That warning arrives as businesses increasingly connect AI services and workplace applications to internal data. The more integrations a business enables, the more important it becomes to understand what is running behind the interface.

Why it matters for businesses

A small organisation may not maintain its own software libraries, but it can still be affected by a supplier incident, a malicious update or a compromised integration. It may also have limited visibility into how a provider protects data, detects suspicious activity and communicates during an outage.

This is a business continuity issue as much as a technical one. If a critical SaaS provider becomes unavailable, changes its terms or suffers a breach, the organisation needs a practical alternative. The cost of preparation is usually lower than the cost of discovering a dependency during an incident.

Practical recommendations

• Maintain a current list of critical suppliers, applications, integrations and data flows.
• Classify suppliers according to the sensitivity of the data and the importance of the service.
• Check whether contracts cover breach notification, data location, access logging, backup arrangements and exit assistance.
• Restrict third-party integrations to the permissions they genuinely require and review them after major changes.
• Test a manual or alternative process for the systems that would stop the business from operating.

Supplier assurance does not require a large procurement department. A simple risk-ranked register and an annual review can expose gaps that would otherwise remain invisible.

What This Means For Your Business

This week’s technology news is not a reason to avoid AI or modern collaboration tools. It is a reason to introduce them deliberately. The strongest business case for AI is not based on excitement about a new feature. It is based on a clearly defined problem, measurable time savings and controls that protect customers, staff and the organisation.

Start by reviewing the identities and integrations that already exist in your business. Confirm that MFA is enabled, privileged access is limited, external collaboration is controlled and important data can be recovered. Then choose one practical AI or workflow improvement and run it inside a documented pilot.

For Australian businesses, the most useful approach is steady improvement rather than a large technology reset. Make secure behaviour part of normal work, give employees a trusted way to verify support requests and ensure every automated action has an owner. That combination will help your organisation capture productivity gains without losing visibility or control.

Call Webwire on 08 9386 0053 or contact us at enquiries@webwire.com.au.