Agentic AI in Lead Generation: The Evolution of Cold Email Outreach

27 August 2026

By Rikesh Vekaria and the Mimecast Threat Research Team

Campaign Overview

Cold email is unsolicited outreach to people who have not asked to hear from the sender, typically a short business-to-business pitch aimed at a work inbox. Used responsibly, it is an inexpensive and measurable way to reach prospective buyers, enter new markets and build outbound pipeline. Used poorly, it becomes inbox noise: generic messages, purchased lists and repeated follow-ups that ignore a recipient's lack of interest or request to opt out.

The ecosystem has changed as generative AI has given way to agentic AI. Generative systems accelerate the writing step. Agentic systems can work toward a standing goal, use connected tools, observe results and continue the outreach loop with limited supervision. Prospect selection, enrichment, personalization, sender rotation, warmup, scheduling, reply handling and meeting booking can now operate as parts of one automated workflow.

Mimecast reviewed two examples of this shift. Instantly.ai combines large-scale cold-email infrastructure with dedicated sales and reply agents. GoZen.io presents a broader go-to-market suite in which agent building, visitor intelligence and cold outreach sit in one stack. These services are not inherently malicious, but they demonstrate how quickly legitimate lead-generation activity can scale and how difficult it can become for end users and security teams to separate relevant outreach from persistent unwanted mail.

From Copilots to Agents

The practical distinction is not simply whether AI writes an email. It is whether the system can pursue an outcome across multiple tools and stages.

For cold email, the bottleneck is rarely a single message. It is the complete loop: deciding whom to contact, finding and enriching the record, choosing what to say, following up, interpreting replies and acting when intent appears. A platform that only drafts text leaves most of this work to a human. An agentic platform attempts to operate the loop.

Campaign Flow

Stage 1: Context and goal intake:The operator supplies a product, website, target market or standing goal such as generating qualified leads.

Stage 2: Lead discovery:The system finds prospects using firmographic, role, database, visitor or intent-style signals.

Stage 3: Enrichment and personalization:Contact records are enriched and messaging is adapted to the recipient, organization, role or recent activity.

Stage 4: Campaign preparation and delivery:The platform creates a sequence, schedules follow-ups and may rotate sending accounts while warmup features support sender reputation.

Stage 5: Reply handling: An agent can classify replies, answer common questions, apply configured tone and product guidance, share a calendar link or escalate when confidence is low.

Stage 6: Continuous feedback: Replies, engagement and campaign outcomes feed the next action. The human role moves from writing every message to approving, supervising and tuning the system.

Instantly.ai

Instantly.ai developed as cold-email infrastructure offering multiple sending accounts, warmup, sequences and a unified inbox, then added goal-seeking agents. Its agentic layer is therefore connected directly to the delivery stack.

AI Sales Agent: Prospect to Campaign

• Context intake: A website URL can be used to infer the product, positioning and likely ideal customer profile.
• Lead discovery: Prospects can be identified through firmographic, role and intent-style signals connected to the platform's B2B contact and SuperSearch ecosystem.
• Enrichment and sequencing: The workflow verifies contacts, drafts personalized multi-step outreach and proposes campaigns.
• Autonomy controls: Human-in-the-loop mode supports approval of findings and sends, while Autopilot can continue without step-by-step approval.

AI Reply Agent: Closing the Loop

The AI Reply Agent treats inbound responses as an always-on workflow. The source material describes rapid handling marketed as under five minutes, configured tone and objection handling, calendar sharing, meeting-oriented responses, low-confidence escalation, and visibility through Unibox, Slack and the Activity Feed.

Together, the Sales Agent and Reply Agent approximate an automated outbound sales-development workflow. Time to first campaign can shrink from manual list building and writing to URL, lead selection and generated copy, while reply handling moves toward machine speed.

GoZen.io

GoZen approaches outreach as part of a unified marketing, sales and support suite. Its model combines specialized products with a general agent builder.

DeepAgent: Autonomous Outreach Workflows

GoZen DeepAgent is presented as an AI agent studio in which users describe goals in natural language or begin with pre-built agents across sales, marketing, support and ecommerce. For outreach, teams can compose multi-step flows spanning capture, qualification, assignment and messaging, and run recurring goals rather than launching only a static campaign.

Growth and HyperReach: Visitor-Led Outreach

• GoZen Growth: Cold outreach and email marketing with sequences, sender rotation, AI-assisted drafting and AI-assisted warmup against a pool of real accounts.
• HyperReach: Company- and people-level visitor identification, generative personalization using intent and LinkedIn-style signals, copy variants with scoring or readability feedback, contact enrichment, and export to CRM or email tools.

This model shifts prospecting from a static list toward a workflow that can notice recent demand signals, identify and enrich the visitor, prepare personalized variants and move the contact into outreach.

What Mimecast Observed

The Instantly.ai figure illustrates the delivery scale available to agentic outreach platforms. Based on customer reports Mimecast blocked over 290 million emails

For GoZen-related outreach, the source dataset contains 443 domain entries representing 442 unique domains. The breadth of the domain set shows how sender infrastructure can rotate across a short observation window. Domain presence in this dataset is not, by itself, proof of malicious ownership or activity.

Recurring GoZen Sender Identity

Mimecast Threat Researchers have identified multiple GoZen outreach messages with the name Ambi Moorthy. A sample message is signed “Ambi Moorthy, CEO | GoZen.io,” and based on a LinkedIn profile it appears that Ambi Moorthy is the CEO of GoZen.io. The following observed From addresses also reuse Ambi or Moorthy in the local part:

Observed sender address

ambimoorthy@gozenflowx[.]xyz
moorthy@getgrowthsaashq[.]com
ambimoorthy@inbound-lab[.]tech
ambi@content-hq[.]store

The repeated use of the executive's name creates continuity of sender identity even when the domain changes. This supports correlation of the outreach cluster, but it does not establish that the named executive personally wrote, approved or transmitted every automated message.

 

Data Protection and GDPR Considerations

Agentic outreach discovers leads, enriches profiles, personalizes content and may answer autonomously, all of which can involve processing personal data. The source research notes that both Instantly and GoZen publish GDPR-oriented policies or contractual materials, but neither service can make a customer's campaign automatically compliant. The customer generally remains responsible for the campaign purpose, lawful basis and use of data.

Shared Compliance Building Blocks

• Lawful basis: For B2B outreach this may be legitimate interests supported by a documented assessment, or consent where required, particularly in consumer or ePrivacy contexts.
• Transparency and rights: The sender should identify itself clearly, provide an effective opt-out and support access, erasure and objection requests.
• Vendor contract: A Data Processing Addendum is needed where a vendor processes personal data on the customer's behalf.
• International transfers: Appropriate safeguards, such as Standard Contractual Clauses, are needed where applicable.
• Operational controls: Suppression lists, unsubscribe handling, minimization, security, retention limits and auditable human oversight become more important as automation increases.

Platform Responsibilities Described in the Source

The source states that Instantly, operated by Foo Monk LLC, publishes a Data Processing Addendum and positions itself as a processor for subscriber campaign data, while customers retain responsibility for a valid lawful basis and campaign operation. Product controls discussed in the source include unsubscribe and suppression handling, list verification, Activity Feed or Unibox auditability, and a choice between human approval and Autopilot.

The source states that GoZen distinguishes between acting as a processor for customer-uploaded outreach data and as a controller for its own account and business data. It also describes a DPA, support for data-subject rights and Standard Contractual Clauses. For Growth, HyperReach and DeepAgent, customers still need purpose limitation, proportionate use of visitor information, opt-out enforcement, retention controls and human escalation for sensitive conversations.

Mimecast Protection

Mimecast telemetry provides visibility into the scale and infrastructure patterns associated with agentic outreach. More than 290 million messages using Instantly.ai were blocked based on customer reports over the past year, while the GoZen-related observation identified more than 440 domains over two months. These patterns can support clustering, policy decisions and unwanted-mail investigation.

Because these platforms have legitimate business uses, platform or domain association should not be the only enforcement signal. Content, sender identity, domain age and behavior, volume, recipient engagement, unsubscribe handling and organizational policy should be evaluated together.

Indicators and Observed Infrastructure

The addresses below are useful for correlation with the described GoZen outreach. They are not presented as confirmed malicious indicators.

The complete source domain set is retained in Appendix A.

Recommendations

User Awareness

• Educate users that a highly personalized message can still be automated and unsolicited
• Encourage recipients to verify the sender, domain and business relevance before replying or following links
• Use unsubscribe or internal reporting mechanisms where appropriate, and escalate messages that ignore an opt-out or use misleading identity claims
• Avoid treating personalization, a familiar executive name or a reference to public social activity as proof of an existing relationship

Conclusion

Lead generation remains a normal and necessary business activity, but agentic AI changes its scale, speed and persistence. The workflow no longer stops at drafting an email: agents can discover prospects, enrich records, personalize messages, rotate senders, manage sequences, respond and book meetings in a continuous loop.

Mimecast's observation of more than 290 million unwanted Instantly.ai messages over one year, alongside more than 440 domains in a two-month GoZen-related dataset, shows the operational reach of this ecosystem. That scale does not make every message malicious, but it increases the burden on recipients, compliance teams and email defenses. Organizations need contextual detection, effective opt-out enforcement, transparent identity and human oversight so that legitimate lead generation does not become persistent unwanted mail.

Appendix A: Observed GoZen-Related Domain Set

The following were observed in the GoZen-related outreach dataset and should not be treated as confirmed malicious solely because they appear here.