Agent Bricks in Motion: Smarter Provide Concentrating on for Gross sales Groups

From Gross sales Dilemma to Knowledge-Pushed Motion

Even the very best industrial presents are solely as efficient as their supply. At Databricks, we offer free credit score presents to assist clients get began or speed up adoption, however gross sales representatives face a deceptively easy query: which of my buyer accounts are eligible, and which ought to I attain out to first?

What looks like a simple process may be opaque and shortly flip right into a time-consuming, multi-team effort, particularly when accounts are unexpectedly ineligible for presents. Gross sales groups usually must dig by means of documentation, seek the advice of Slack threads, and manually examine accounts with operations groups. This creates pointless back-and-forth, slows down momentum, and will get in the best way of offering clients with high-value presents. Even when accounts are identified to be eligible, it’s not all the time apparent which ought to be prioritized.

Constructing a Smarter System with Agent Bricks

To sort out the issue, our staff turned to Agent Bricks — Databricks’ platform for constructing high-quality AI brokers on enterprise knowledge — and constructed a multi-agent system that delivers clear, actionable steering on to gross sales groups. In lower than two days, I created a software that lets gross sales reps:

• Shortly determine which buyer accounts qualify for credit score presents
• Perceive the precise cause an account isn’t eligible
• Rank eligible accounts to give attention to the highest-impact prospects first

As an intern in Enterprise Technique and Operations this summer time, I had a brief turnaround time, so velocity and ease had been crucial. Agent Bricks let me shortly construct a high-quality answer and supply the enablement gross sales groups wanted.

Designing the Multi-Agent Resolution

Utilizing Agent Bricks’ Multi-Agent Supervisor, I designed a system that chains collectively three purpose-built brokers beneath one supervisor. Like an air-traffic controller, the Supervisor decides which agent to delegate every a part of the query to after which stitches their responses into one clear reply.

One Supervisor, Three Specialised Brokers

My answer makes use of three brokers: two AI/BI Genie brokers and a Data Assistant agent, managed by a supervisor to orchestrate duties and knowledge movement:

While not having to analysis supervisor agent structure or have interaction with technical groups, I used to be in a position to construct a practical AI agent system immediately on our buyer knowledge and supply program paperwork.

From Handbook Requests to Self-Serve Insights

The multi-agent answer removes guesswork and creates a seamless, explainable expertise. By combining structured buyer knowledge with unstructured supply program info, the system permits:

• Self-serve eligibility troubleshooting: As a substitute of routing by means of a number of groups and Slack threads, gross sales groups can now shortly perceive supply eligibility points and take knowledgeable motion immediately, because of built-in explanations
• Extra clever focusing on: Gross sales groups can give attention to high-value accounts based mostly on actual progress alerts and supply relevance, not hunches, streamlining how they determine high-impact alternatives
• Sooner outreach: By growing supply understandability and lowering guide friction, the response SLA decreases from 48 hours to beneath 5 seconds, and gross sales groups can transfer extra shortly and confidently

Most significantly, the system scales as accounts are added and extra presents are created. Buyer account and GTM insights replace robotically when the reference knowledge in Unity Catalog adjustments, and new supply applications may be supported by updating the paperwork within the data base – with no new code required.

Limitations

Whereas the present system is highly effective, there are just a few limitations to notice:

• Agent Overlap: As a result of the brokers can’t immediately share context, sure items of knowledge wanted to be duplicated throughout them, regardless that the supervisor “is aware of all.” For instance, the Account Prioritization agent’s knowledge desk features a column indicating which supply – if any – every account is eligible for (already identified to the Eligibility agent). It additionally comprises context concerning the utilization eligibility bands for every supply kind (already identified to the Provide Particulars agent). This duplication ensures the Prioritization agent can cause about focusing on and rank accounts accurately.
• Consumer Workflow Integration: Most gross sales groups work primarily in Slack and Salesforce, not Databricks. Integrating this technique as a Slackbot or into Salesforce would put eligibility particulars and steering immediately into their on a regular basis workflows.

Conclusion

Industrial presents solely work if gross sales groups know who to focus on — and why. Earlier than Agent Bricks, this was a guide, multi-team problem that slowed down outreach and launched ambiguity into our applications. With Agent Bricks, we had been in a position to construct, take a look at, and refine a multi-agent AI system with nothing extra in hand than our knowledge and our aim.

Although our system has just a few limitations in its present type and isn’t embedded within the instruments gross sales groups use day by day, the beneficial properties have already been significant; it’s made supply focusing on sooner, extra clear, and extra scalable. The actual magic lies within the prioritization of accounts: the system robotically aggregates buyer knowledge and supply info to intelligently floor the highest-impact alternatives first, and I didn’t even have to inform the agent precisely learn how to do it. Now that’s knowledge intelligence. 
 

Get began constructing with Agent Bricks and create your first answer at the moment.