When your AI agent guesses what revenue means and lives in one person's chat window, the ROI you promised never shows up. That gap was the thread running through our recent AgentSpot: Multiplayer AI in Action event.
I had the pleasure of hearing Francois Lopitaux, Nicolas Rentz, and Kelley Jarrett of ThoughtSpot walk through how shared agents close it.
We were also joined by Pedro Mayoral Picazo, Senior BI Manager at Enmacc, Europe's leading over-the-counter trading platform for energy and environmental commodities. What did his BI team get out of AgentSpot?
Here’s how orgs just like yours are approaching multiplayer AI, and what you can immediately put to work for your own team.
How to Close the Context Gap Before You Scale Agents
Francois Lopitaux, SVP of Product Management at ThoughtSpot, opened with two problems we suspect you'll recognize. How often do you still share AI output through emails and screenshots, the way people once emailed Word doc versions back and forth?
The second problem stuck with us more: AI reads your data without understanding it.
Picture a column called rev_c, which could mean recognized or contracted revenue, on a fiscal or calendar basis. Your AI has no way to know, so it guesses.
To close that gap, Francois walked us through five principles:
Multiplayer AI
Memory
Knowing your business context
Governance
Cost
Each one showed up in how AgentSpot handles the basics:
Shared Definitions: AgentSpot pulls revenue, accounts receivable, and fiscal calendar from the semantic layer, so finance, sales, and marketing see the same numbers.
Full Visibility: IT sees every agent, workflow, and app with audit logs, so shadow AI stays in check.
Inherited Permissions: Every run uses the requesting user's permissions, so agents never see more data than the person asking.
Outcome Pricing: AgentSpot pricing covers unlimited users and agents, with one credit per conversation, workflow run, or app.
Why does that last point matter so much? With no per-token billing, you can forecast spend before your first rollout.
"Whatever be the complexity, whatever be the model, we keep it super simple so you can predict what it is going to cost you," Francois explained.
Shared Agents in Four to Five Minutes: How to Build by Conversation
With your definitions and costs settled, the role of the builder can shift as well. Nicolas Rentz, Senior Director of Product Management at ThoughtSpot, showed what it looks like when anyone on your team can build.
Ever fired a one-line prompt at a chatbot and gotten a guess back? "When you ask a simple question to a chatbot, it doesn't really end well," Nicolas said.
The fix starts with Scout: an entry point that routes your request to an agent a colleague already built, pulling from Salesforce, Slack, Jira, and Google Workspace.
But what if no agent exists yet? You describe the job in chat, and AgentSpot writes the objective, audience, allowed tools, workflow, and output format.
From there, you have:
Shared Memory: Organizational memory stops builders from redefining your fiscal year, and personal memory tailors shared agents to each user.
Multiplayer Threads: You can pull a teammate into a multiplayer AI thread, or text an agent when you're on the go.
Context-Aware Slack: My favorite moment was the agent ignoring karaoke chatter and jumping in when someone asked about an account's temperature.
Fast Workflows: Nicolas built a morning meeting-prep agentic workflow by chat in four to five minutes and scheduled it for 7:10 a.m.
"You don't have to be an architect to figure out how to create a deterministic workflow with AgentSpot. You just have to chat with us," Nicolas added.
Weeks of Work in One Day: How Enmacc Connected Data Without a Three-Month Project
Building by chat pays off fastest when your data sits in silos, which is exactly where Enmacc started. ThoughtSpot is the main BI platform for Pedro's team, but they had barely used AI beyond scheduling meetings and pulling spreadsheet data.
So what was in the way? Their customer relationship management (CRM) data sat outside the warehouse.
"So before we're planning to do a huge project for three months and to see how we can embed or how to connect these two different sources," Pedro explained.
Instead of waiting three months, Pedro and his team tried AgentSpot. Here’s how the work broke down:
The Data Gap: With CRM data outside the warehouse, the team faced a three-month project before anyone could blend the two sources.
The Manual Burden: No central wiki existed for stakeholders, so every documentation update took hours of manual effort.
The AgentSpot Fix: Pedro mixed sources with nothing implemented in the background, then built a Google Site fed entirely by an agent via API.
The Payoff: Work he expected to take a couple of weeks got done in one day, and tables now get documented in one click.
How fast is maintenance now? Once the agent learned to document, each update dropped to about two minutes.
"Now every time that we have to update any type of documentation or we have new libraries and so on, it takes us maybe two minutes to do it," Pedro said.
Eight Hours Back per Seller, per Week: How to Turn Seller-Built Agents into a Governed Hub
Once agents spread beyond one team, ownership gets messy fast. Kelley Jarrett, SVP of Strategy, Operations, and Enablement at ThoughtSpot, described exactly that inside our own sales org.
"As these agents were built before AgentSpot, they were living and dying with the person that was building them," Kelley explained.
Who paid the price? Our sellers did, trading selling time for building time on agents that weren't shared or connected to core systems. Kelley's team answered with an AI sales hub on AgentSpot:
The Territory App: Sellers see priority accounts, intent signals, buying-group changes, and new leaders in one shared view.
Account Research in Slack: An agent pulls positioning and competitive advantages from the content management system without sellers leaving Slack.
Deal Progression: A separate agent draws on the sales process checklist, so every seller works from the same checklist.
Central Governance: Some of these sales agents started with non-technical team members and sellers before Kelley's team connected and governed them.
"We built an AI sales hub using AgentSpot ... that saves our sellers eight hours per seller per week," Kelley said.
Because the agents now live in one governed hub, they no longer depend on whoever built them.
Advice for Building Your First Agent: 4 Practical Tips
If you're wondering where to start, Pedro's experience offers a grounded path. Here is what I'd take from his approach:
Write the Prompt Yourself: Pedro was candid that the work isn't automatic, so budget time to write and refine the prompt.
Start Small: Enmacc began inside its BI team and now wants faster stakeholder access to data so it can impact sales.
Build the Next Agent: His team is still building a sales agent that would return 25 key performance indicators for any company name.
Plan Ahead: Pedro's next goal is predictive analytics that mixes external market data with internal data every day.
How to Build Your First Governed, Shareable Agent
What stuck with me most after the session: agents pay off when they share your definitions, your permissions, and your teammates. Why does that matter for your ROI? Shared, governed agents keep paying off after their builder moves on.
AgentSpot is available now - See it in action and build an agentic workflow in 30 minutes.
