Abacum embedded a Guru Knowledge Agent directly in Slack to answer product questions for its globally distributed post-sales team. As a result, new hires in complex roles ramped up in a third of the time, and product expertise was no longer locked up in a small group of specialists in Barcelona.

"Having the Knowledge Agent in Slack encouraged people to be more curious, because the information isn't being gatekept as much as it was previously."

— Baha Makarem, Head of Professional Services, Abacum

The Challenge: Tacit Knowledge, Global Team

Abacum is an AI-native FP&A platform that helps finance teams replace manual, spreadsheet-heavy planning with automated forecasting, budgeting, and reporting. Founded in 2020, the company has grown to roughly 180 employees serving finance teams around the world, with core product and implementation expertise concentrated in Barcelona and a post-sales organization spread across time zones and languages.

Experts in One City, Questions Everywhere Else

Barcelona was where Abacum's deepest product knowledge lived. Teammates in other regions who needed an answer to a specific implementation question had to track down one of a handful of experts and wait, which slowed down delivery timelines and made it hard for newer hires outside Barcelona to build confidence quickly.

Tacit Knowledge, Locked in Individuals

Much of what the post-sales team needed to know didn't exist in a document anywhere. It lived in the heads of a few people, surfaced only when someone happened to ask the right person the right question in Slack or on a call.

"I viewed the problem as: how do we make our distributed team work together better, versus having single people be the fallback."

— Arpan Patel, AI Finance Solutions, Abacum

A Product That Wouldn't Hold Still

Abacum's platform changes constantly, and static documentation couldn't keep pace. Whatever guidance existed tended to lag behind the product itself, so even motivated new hires were often learning from information that was already out of date.

The Approach: Connect, Access, Continuously Improve

Abacum built a single Guru-powered knowledge layer for its post-sales team, then made sure it stayed both easy to reach and accurate over time.

1. Connect: One Agent, Deliberately Curated Sources

Arpan Patel, Abacum's AI Finance Solutions expert and Guru champion, connected the knowledge agent to a specific mix of sources: curated Slack channels, Guru itself, Notion for historical context (via MCP), and select external sites, like Fivetran and NetSuite documentation, for integration details Abacum doesn't maintain internally. The curation was deliberate. Rather than connecting every available channel and document, Arpan scoped the knowledge agent to sources that were actually relevant, keeping irrelevant chatter and noise out of the knowledge base entirely.

2. Access: Answers Inside the Channel Where Questions Already Happen

Instead of asking the team to go somewhere new for answers, Abacum built the knowledge agent into a Slack channel called "Abacum Overflow," where anyone can ask a product question and get an answer in-thread. When an expert corrects a wrong answer, that correction is captured and feeds back into the agent's source material, so the fix compounds over time instead of getting lost in a Slack thread. What started as a tool for the delivery team spread organically to customer success, and now sales reps use it too.

"If the answer is wrong, someone comes in and corrects it. And the great thing is, that's saved for eternity — the next time somebody asks that question, it knows the card was wrong."

— Arpan Patel, AI Finance Solutions, Abacum

3. Continuously Improve: Closing Gaps Automatically

A weekly automation, built in Claude and connected to Guru through MCP, scans Abacum's slack channels used to answer questions and compares them against what the Guru agent was actually able to answer. Every question flagged as a high-risk knowledge gap gets auto-drafted into a new Guru card. Arpan reviews each draft and publishes it with a single click. This approval flow means no card gets built manually from scratch, saving Arpan tons of time allowing him to close gaps at a rate that would not otherwise be possible.

The result is a knowledge base that repairs its own gaps on a weekly cadence, without turning card creation into a manual chore for any one person.

The Results

• 🚀 New-hire ramp time reduced as much by 60% — one junior hire, working outside Barcelona, reached a proficiency level in four months that previously took new employees about a year to hit.
• 📈 Right answers, when and where they need it - Thousands of questions answered by Guru in Slack within 30 days. 71% of Abacum's team asking Guru at least one question in that window.
• 📋 Hundreds of knowledge gaps closed – Guru cards built from automations, human-approved. Sourced from dozens of connected sources.
• 🔥 Adoption spread organically from the delivery team to customer success to sales reps, without a formal companywide rollout push.

Beyond the numbers, the bigger shift has been behavioral. Most of Abacum's current post-sales hires have been on the job less than six months and are now performing at a level that used to take much longer to reach, without needing to interrupt a product manager or wait for a specific expert to be available.

Delivery, customer success, and sales all draw from the same governed source of truth, so an answer verified once in Barcelona is available everywhere the team works.

That shift is what they pointed to as a real change: people aren't just getting faster answers, they're asking more questions in the first place, because trusted information is no longer living in the not-easily-accessible heads of specialists.

What's Next

As Abacum continues to grow, Guru is set up to scale along with it. The same curated-source, self-correcting knowledge base that ramped up that junior hire can onboard the next hire, and the one after that, without adding more strain on Barcelona's experts. Product complexity and offerings will continue to change, but the corrective feedback loop is now solidly in place to ensure accurate information always finds people wherever they work, whenever they need it.