Hightouch Appoints Deepscribe Cofounder Akilesh Bapu to Lead AI Product Development for Autonomous, Context-Aware Marketing Agents

Akilesh Bapu Joins Hightouch to Lead Generative AI Product Development

Hightouch has appointed Akilesh Bapu, co-founder and former CEO of Deepscribe, as Head of AI Product Development. Bapu scaled Deepscribe into a high-growth healthcare AI company serving thousands of clinicians with ambient scribe technology, led multiple funding rounds, and built its core AI platform.

His mandate at Hightouch: build autonomous, context-aware marketing agents on top of the company's Composable CDP foundation. With deep integrations into enterprise data platforms, Hightouch is positioning AI agents to act with high-quality context and measurable feedback loops.

Why this matters for product development

Marketing is a closed-loop system: every action can be tied to data and outcomes. That makes it an ideal proving ground for agentic AI that can plan, execute, and learn within guardrails.

Hightouch's integrations with platforms like Snowflake and Databricks give agents direct access to enterprise-grade context. For product teams, this reduces the gap between model reasoning and real business state, which is where most AI products falter.

What Bapu brings

Bapu built and shipped applied AI in a regulated domain, balancing product velocity with accuracy and reliability. He grew Deepscribe through partnerships with major health systems, showing a pragmatic approach to distribution and trust.

"Akilesh's track record of building transformative AI products and his conviction around the future of agentic workflows make him an extraordinary addition to the team," said Tejas Manohar, co-founder and CEO of Hightouch.

"As foundation models commoditize, the real breakthroughs will emerge at the application layer," said Akilesh Bapu. "Hightouch's combination of deep enterprise data access and agentic reasoning creates the conditions for AI marketing agents that assist and also operate autonomously to drive measurable outcomes."

Product strategy signals to watch

• Closed-loop agents: Agents that own a metric (e.g., ROAS, CAC, LTV), run experiments, and self-correct based on warehouse truth.
• Context-first design: Retrieval over the warehouse/CDP to ground prompts, plans, and actions in customer, campaign, and spend data.
• Action interfaces: Safe execution layers across ad platforms, email/SMS, and CRM with fine-grained permissions and auditability.
• Evaluation and guardrails: Offline and online evals, policy checks, and rollback paths for agent actions.
• LLMOps integration: Versioned prompts, datasets, simulators, and live monitors tied to business KPIs.
• Data governance: PII handling, consent, and compliance embedded into agent capabilities from day one.

Implementation checklist for product teams

• Define the success metric your agent will own and the boundaries of its control (channels, budgets, frequency, audiences).
• Centralize truth in the warehouse; standardize schemas for customers, events, spend, and outcomes.
• Build retrieval pipelines that enrich prompts with current state, constraints, and recent performance.
• Create a policy layer: allow/deny lists, budget caps, approval thresholds, and audit logs.
• Start with human-in-the-loop: propose actions, then graduate to autonomous execution with auto-revert on anomaly.
• Ship an evaluation harness: offline replay, counterfactuals where possible, and online A/B or multi-armed bandits.
• Monitor drift and data quality; alert on broken connectors, missing contexts, and KPI deviations.
• Document failure modes and escalation paths; test chaos scenarios before enabling full autonomy.

Stack considerations

Use your warehouse as the context spine; keep features, audiences, and outcomes in one place. Connect agents to execution systems with strict scopes and reversible actions.

Favor interpretable plans and structured action outputs over free-form text. Treat prompts, tools, and policies as versioned artifacts so you can roll forward and back without drama.

What to expect next

Hightouch is likely to ship agents that can manage segments, budgets, creatives, and channel mix with tight feedback loops to warehouse metrics. The focus will be measurable lift, safer autonomy, and faster iteration across channels.

Helpful resources

• AI Automation Certification - practical frameworks for building and governing agent workflows.