ON-DEMAND WEBINAR
Strategies for Scaling Agentforce AI Systems
Published on March 7, 2026
Plenty of Agentforce projects stall after the first agent goes live. In this session hosted by Salesforce Ben, Diabsolut and Salesforce share what it takes to move AI agents from pilot to production: setting a realistic success baseline, fixing the knowledge gaps agents expose, and running a steady refinement cycle after launch. It’s for Salesforce leaders, admins and architects who are planning or already running Agentforce.
What you’ll learn
- How to judge an AI agent: Measure it the way you would a new hire, by the volume of work it handles, the quality of that work and whether it keeps improving.
- The institutional knowledge problem: Why agents fail on processes nobody wrote down, and how to treat knowledge as a product with an owner and a maintenance cadence.
- Agentforce compared with ChatGPT: Why a grounded orchestration layer behaves differently from a consumer chatbot, and what that means for liability.
- The 40/60 implementation model: Spend 40% of the effort building and 60% refining, with a regular review that sorts issues into data, agent optimization, user training and technical limits.
ABOUT THE SESSION
Define success with data, then go live and refine
Diabsolut used this approach to take a higher education customer’s service agent from under a 10% good response rate to 82.5%, where support reps wouldn’t change a word of the agent’s reply.
Teams test an agent, see it miss a few answers and decide it isn’t ready. Others spend months in UAT and still face a long cleanup after launch.
Why go-live is the starting line
An AI agent combines an LLM, Salesforce, your data and the people using it, so the list of possible test scenarios is close to endless. The panel explains why reaching production quickly gives you the real-world data you need, and why most organizations reach the results they want within about 90 days of steady monitoring and tuning. Diabsolut now includes three months of AgentGuard, its post-go-live monitoring and optimization service, with every Agentforce deployment.
The work after go-live is where enterprise AI earns trust.
SPEAKERS
Meet the expert panelists
Christine Marshall
Moderator/Author, Salesforce Ben
Christine is a 12x certified Salesforce Hall of Fame MVP and leads the Bristol Admin User Group.
Shannon Jensen
Success Architect, Salesforce
Shannon is an Agentforce specialist with four years at Salesforce and a background building data pipelines in digital advertising.
John Pettifor
SVP, Innovation, Diabsolut
John has spent fifteen years in the Salesforce ecosystem as a customer, partner and in customer success at Salesforce before joining Diabsolut.
Zain Ladha
Director, Outcomes Consulting, Diabsolut
Zain helps companies plan, implement and release AI, including Agentforce, into their operations to reach real business outcomes.
Frequently asked questions
You can build an agent in less than a month. Plan for about 90 days of monitoring, feedback and optimization before it reaches the performance you’re aiming for.
The panel recommends starting with an internal use case, where you have more control over expectations, change management and communication, before launching to customers or partners.
Start with a narrow use case, clean the data it needs, and test it thoroughly. Agents are modular, so you can limit one to specific objects or fields and expand from there.