Building Your Agency AI Intelligence Layer, with Khusbu Doshi –– Special Episode

About this Episode

In this special episode of the Agency Profit Podcast, Kristen Kelly is joined by Khushbu Doshi, COO at E2M and AI integration specialist, to explore how agencies can move beyond experimenting with AI and start building an intelligence layer that makes their systems smarter over time. Drawing from her experience working with hundreds of agencies and helping E2M scale its own AI capabilities, Khushbu explains why agents and automations become far more useful when they have access to the context, history, and institutional knowledge that experienced team members naturally bring to their work. Together, Kristen and Khushbu break down the progression from context-aware AI agents to self-learning automations and custom AI systems that connect information across clients, teams, sales, finance, and delivery. From creating and maintaining client context files to identifying opportunities, improving retention, refining delivery, and connecting data across the agency, this episode offers a practical roadmap for turning scattered AI experiments into systems that continuously learn and help teams make better decisions.

Points of Interest

• 00:01 – 03:16 – From AI Experimentation to Transformation: Kristen introduces Khushbu Doshi and explores why many agencies are still struggling to turn rapidly evolving AI tools into practical, repeatable improvements.
• 03:16 – 06:14 – Why AI Systems Need to Keep Evolving: Khushbu explains that building an AI solution once is not enough because new models and capabilities can dramatically improve the efficiency of existing workflows over time.
• 06:19 – 08:25 – Giving AI Institutional Knowledge: Khushbu compares an AI agent without company context to a new intern, showing why accumulated knowledge about clients, processes, and past decisions can dramatically improve AI output.
• 08:25 – 11:46 – Defining the AI Intelligence Layer: Khushbu introduces the intelligence layer as the connective context that sits across an agency's AI agents, automations, applications, and organizational knowledge.
• 11:46 – 15:15 – Context Changes the Quality of AI Output: A client example demonstrates how the same AI agent and playbook can produce significantly more relevant recommendations when given information about competitors, previous work, performance, and client preferences.
• 15:15 – 17:55 – What Belongs in a Client Context File: Khushbu outlines the information AI needs to make better decisions, including client profiles, competitive landscapes, historical work, rejected ideas, comparable outcomes, market dynamics, performance, and strategic notes.
• 17:55 – 23:19 – Agents, Automations, and Self-Learning Workflows: The conversation distinguishes individual AI agents from connected automations and explains how memory can allow workflows to learn from meetings, objections, outcomes, and other recurring inputs.
• 23:19 – 28:28 – Building an Agency-Wide AI System: Khushbu explains how custom AI applications can bring together information from finance, project management, CRM, sales, meetings, and other sources while an intelligence layer identifies relationships across that data.
• 28:28 – 32:42 – Turning Agency Data Into Actionable Decisions: Khushbu demonstrates how an agency brain can identify opportunities across clients, recommend priorities, match opportunities with the right account managers, and explain the reasoning behind those recommendations.
• 32:42 – 36:39 – A Practical Roadmap for Building the Intelligence Layer: Agencies can begin with regularly updated client context files, progress to context-aware agents and self-learning automations, and eventually build relational databases and custom AI infrastructure as their needs become more sophisticated.
• 36:39 – 39:46 – Applying Intelligence Across Sales, Retention, and Delivery: Khushbu shares how intelligence layers can improve lead qualification, sales, retention, employee training, and client delivery by connecting historical outcomes with current decisions.
• 39:46 – 44:56 – Cost, Implementation, and Getting Started: Khushbu discusses the investment required to build and maintain AI infrastructure and explains how E2M helps agencies identify high-value AI opportunities, deploy solutions, train teams, and eventually offer AI capabilities to their own clients.

Show Notes

• Connect with Khusbu Doshi
• E2M Solutions
• E2M Guides
• Vistara AI for Agencies on Feb 24-26, 2027 at Hilton Bayfront, San Diego, California