Multi-channel social engineering now outpaces manual defence

At a glance

What this is: Doppel argues that social engineering is becoming multi-channel and AI-native, with attacks now spanning collaboration tools, social platforms and live meetings rather than email alone.

Why it matters: That shift matters because identity teams now have to govern trust, verification and response across human and non-human touchpoints, not just mail gateways and phishing filters.

Context

Social engineering is no longer a single-channel problem. When attackers can move from email into collaboration platforms, social media and live meetings, the weak point is no longer only message filtering but the organisation's ability to verify intent across channels.

The security gap is governance, not just detection. Traditional stacks are built for after-the-fact review, while AI-assisted deception compresses the time available for human judgment and manual investigation. That makes collaboration-tool oversight and identity verification controls part of the same defense model.

Key questions

Q: What breaks when phishing moves beyond email into collaboration apps?

A: Email-only defence breaks because the attacker no longer needs to win inside the mailbox. They can exploit trust in chat, social or file-sharing channels where users are less primed for scrutiny and where identity cues are weaker. Teams need verification, reporting and monitoring that follow the user across channels, not just the inbox.

Q: Why do AI-generated impersonation attacks make manual investigation less effective?

A: AI reduces the cost of creating believable, varied lures at scale, so analysts face more incidents that look plausible and arrive faster than they can review them. Manual investigation becomes a bottleneck when every case needs separate human validation. The answer is correlation and automation first, with human judgment reserved for exceptions and high-impact decisions.

Q: How should security teams handle social engineering when launch happens across multiple channels at once?

A: Teams should treat launch as a campaign correlation problem, not a single-message problem. The right response is to connect email, SMS, voice, social and SaaS telemetry so the operation is visible as one staged event. That lets defenders disrupt the infrastructure behind the lures instead of only removing individual messages as they appear.

Q: What is the difference between phishing detection and social engineering defense?

A: Phishing detection focuses on identifying malicious messages, usually in email, while social engineering defense covers the broader trust attack that can start in chat, social media, voice or live meetings. The latter requires identity verification, workflow controls and rapid response across channels. If the defence only inspects messages, it misses the persuasion stage.

Technical breakdown

Why multi-channel deception defeats email-centric controls

Email security controls still matter, but they only cover one entry path. Modern social engineering can begin in a direct message, continue in a collaborative workspace, and finish in a live call where the attacker exploits trust built earlier in the chain. The defender is not just checking message content, but trying to verify the sender, the context and the legitimacy of the request across multiple channels. This breaks point-in-time review models because the attack surface is now distributed across systems that do not share the same evidence trail. See the Collaboration Tool Security Guide.

Practical implication: extend verification and monitoring beyond email into collaboration platforms, social channels and meeting workflows.

How AI-driven deception changes the investigation model

AI-native deception reduces the effort needed to scale persuasive impersonation, vary language and sustain many concurrent lures. That changes the economics of investigation: analysts can no longer assume each suspicious message is an isolated event that can be manually triaged in sequence. Instead, responders need correlated signals, rapid containment and workflow automation that can keep pace with adversary volume. The mechanism matters because the attacker is optimising for speed and believability at the same time, which makes human review the bottleneck rather than the cure. See the Deepfakes, Social Engineering and AI Impersonation Guide.

Practical implication: build automated enrichment, correlation and takedown workflows so investigators are not the pacing constraint.

Why agentic defense becomes the operating model

Agentic defense means using machine-assisted workflows to detect, correlate and act across multiple attack channels without waiting for every step to be reviewed by hand. In this context, the agent is not replacing analysts, but absorbing repetitive scanning, clustering and response tasks that are too slow for modern deception campaigns. The architectural shift is from isolated alerts to orchestration across identity, brand and fraud signals, with the human focused on judgment and escalation. That is where identity governance intersects with social engineering: if trust is the target, response has to be identity-aware and machine-speed.

Practical implication: define where automated response can act first and where human approval must remain mandatory.

Threat narrative

Attacker objective: The attacker wants to exploit trust at scale so that victims approve actions, expose access or hand over data without recognising the deception in time.

1. Entry begins in one of several legitimate-looking channels, including email, Slack, Teams, social media or a live meeting, so the attacker can choose the venue that best matches the target's routine.
2. Credential or trust access is achieved by impersonating a known person, brand or workflow, then sustaining the interaction long enough to make the request appear normal.
3. Escalation happens when the target moves from reading the message to acting on it, whether that means approving a payment, sharing access or following a malicious instruction.
4. Impact is the compromise of trust, data or funds, with defenders often learning too late because the interaction looked legitimate at the point of contact.

Breaches seen in the wild

• Uber breach 2022: A contractor's stolen password and MFA fatigue gave a Lapsus$-linked attacker Uber's internal tools; Uber rotated keys to many services.
• SalesBleed Salesforce Agentforce 2026: Three fixed Agentforce flaws let poisoned web leads make AI agents leak CRM data with zero clicks and send phishing under the agent's identity.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Multi-channel social engineering is now an identity governance problem, not just a messaging problem. Once attacks move across Slack, Teams, social media and meetings, the security question becomes whether the organisation can verify who is asking, in which context and with what authority. That is the same trust problem that underpins collaboration tool security and broader identity verification. Teams that still treat phishing as an email issue are defending the wrong layer.

Verification trust gap: the real failure is assuming a legitimate-looking interaction is safe until proven otherwise. AI-native deception exploits the time between first contact and human confirmation, which is exactly where manual review is weakest. The governance gap is not a lack of awareness, but a lack of channel-spanning proof that the person, brand or request is authentic. Practitioners should treat verification as a control plane, not a user behaviour problem.

Agentic defense will become the baseline for high-volume deception response. When attackers can generate, vary and distribute lures at machine speed, the defender must automate detection, correlation and initial containment just to keep pace. That does not eliminate human review, but it changes where human judgment enters the workflow. The practitioners who win this fight will standardise machine-speed response and reserve humans for escalation and exception handling.

Brand abuse, account takeover and fraud are converging into one operational risk class. The article's core point is that modern deception no longer respects the old boundaries between email security, fraud operations and identity governance. That convergence means security leaders need shared telemetry, shared triage criteria and shared ownership across those teams. A siloed operating model will miss the attack until the damage is already done.

Social engineering defense maturity will increasingly be measured by response speed, not alert volume. The important question is no longer how many suspicious messages were flagged, but how quickly the organisation can verify, correlate and neutralise a campaign across channels. That pushes programmes toward orchestration, pre-approved takedown paths and identity-aware verification steps. The organisations that can act in minutes, not hours, will limit trust loss most effectively.

From our research library:

• 28% of secrets incidents now originate outside code repositories, in Slack, Jira, and Confluence, and are 13% more likely to be categorised as critical than code-based leaks, according to the State of Secrets Sprawl 2026.
• Read next: Collaboration Tool Security Guide

What this signals

Channel-spanning trust abuse is the new baseline risk: defenders should expect impersonation attempts to move fluidly between collaboration platforms, social media and live meetings, which means controls must verify context as well as content.

If organisations still measure social engineering readiness by email-filter performance alone, they are undercounting the problem. The more useful indicator is whether a team can correlate a suspicious interaction across channels and act before the request is completed.

The strongest programs will build machine-speed response around identity-aware verification, because the operational contest is now between attacker throughput and defender coordination, not just between good and bad messages.

For practitioners

• Expand monitoring beyond email Instrument Slack, Teams, social channels and meeting workflows so suspicious impersonation is visible where it starts, not only where it lands.
• Build verification steps into high-risk requests Require out-of-band identity checks for payment changes, credential resets, executive instructions and other high-impact requests that attackers commonly impersonate.
• Automate correlation across brand and identity signals Connect collaboration alerts, impersonation reports and fraud indicators so analysts see a campaign rather than isolated messages.
• Pre-authorise containment workflows Define who can disable accounts, remove posts, trigger takedowns or quarantine messages before the incident becomes a prolonged trust event.

Key takeaways

• Social engineering is evolving into a multi-channel trust attack that crosses collaboration tools, social platforms and live meetings, so email-centric controls no longer cover the full threat.
• The shift to AI-driven deception compresses investigation time and makes manual triage an unreliable primary defence at scale.
• Programmes that combine verification, correlation and rapid containment will reduce trust loss faster than teams that rely on detection alone.

Related NHI Mgmt Group resources: Collaboration Tool Security Guide · Workforce Identity Security Guide · Identity Provider and SSO Security Guide · Account Recovery and Help Desk Security Guide

Key terms

• Multi-Channel Social Engineering: Multi-channel social engineering uses more than email to deceive targets, including SMS, voice calls, and impersonation through video or chat. Defending against it requires verification habits and reporting workflows that work across every channel where trust can be manipulated.
• AI-Native Deception: The use of AI systems to create, vary and scale deceptive content and interactions. In practice, this reduces the cost of believable impersonation and shortens the window in which human review can detect that a request is fraudulent.
• Agentic Defense Loop: An agentic defense loop is a continuous sequence in which a security agent finds a weakness, proves the exploit, proposes or creates a fix, and then retests the deployed remediation. The model is valuable because it closes the gap between discovery and validation, but only if access, evidence, and review are controlled.
• Activation Trust Gap: The activation trust gap is the difference between trusting data because it is protected and governing it because it is being reused. It appears when organisations move data from backup or archival systems into AI pipelines without reapplying access, sensitivity, and consumer controls.

Deepen your knowledge

The NHI Foundation Level course, the industry's only accredited NHI security programme, covers NHI governance, machine identity security and identity lifecycle principles that help teams reason about trust and verification. It gives identity and security practitioners a structured way to connect access governance to the wider control stack.