How did Ontop use CustomGPT.ai?
Ontop, a Y Combinator-backed global payroll company, used CustomGPT.ai to build “Barry,” an internal AI knowledge assistant deployed inside Slack. Barry answers sales team questions about international compliance, payroll, and EOR regulations using Ontop’s own documentation, with every response backed by a citation to the source. The result: Ontop’s legal team saved 130 hours per month, response time dropped from 20 minutes to 20 seconds, and the sales team gained the ability to self-serve answers without interrupting legal staff.
Key Results
- 130 hours saved per month legal team time redirected from FAQ responses to strategic work
- Response time cut from 20 minutes to 20 seconds 60x faster
- 400+ complex queries answered monthly by Barry without human involvement
- 60% acceptance rate in a legally sensitive, compliance-critical environment
About Ontop
Ontop is a global payroll and Employee of Record (EOR) services company. Backed by Y Combinator and recognized as one of Colombia’s Top 100 Most Important Startups for 2022, Ontop helps international organizations hire, onboard, and pay employees worldwide while maintaining full legal compliance across each country’s requirements.
With a 200-person team that includes specialized legal experts managing multi-jurisdiction payroll and compliance, Ontop operates in a high-complexity environment where information accuracy is non-negotiable. The initiative to deploy an internal AI agent was led by Tomas Giraldo, Product Manager for the Employee of Record product line.
Ontop’s Challenge
Ontop’s sales team regularly needed answers to questions about compliance rules, payroll requirements, and hiring regulations across international jurisdictions. Although the relevant documentation existed internally, sales reps rarely consulted it, instead defaulting to asking the legal team directly.
The legal team found itself fielding 100+ repetitive questions every week. Each interruption pulled a legal expert away from strategic work to answer questions that, in principle, could have been self-served.
In Tomas Giraldo’s words:
“To reduce the manual processes and operational tasks our legal team was facing when having to constantly answer frequently asked questions. And, so that they could focus entirely on strategic tasks rather than answering questions salespeople could direct to a different place.”
The problem was compounded by the compliance sensitivity of those answers. Informal, verbal responses to payroll or EOR questions given under time pressure created accuracy risk. There was no audit trail, no consistency, and no way to measure how well the legal team’s knowledge was actually reaching the sales floor.
Why Sales and Legal Teams Needed an Internal AI Assistant
Ontop’s challenge is not unique. In organizations where specialized knowledge lives with a small expert team, high-volume question traffic from other departments creates two compounding problems:
For legal teams: Every FAQ answered is time stolen from the work only legal experts can do contract analysis, jurisdiction-specific risk assessment, strategic compliance planning.
For sales teams: Every unanswered question is a deal that stalls. When a sales rep has to wait 20 minutes for a compliance clarification, customer response times slow, and focus shifts away from selling.
An internal AI assistant solves both problems simultaneously: sales gets instant answers, and legal gets their time back. The critical requirement, especially in legally sensitive domains, is that those answers are accurate, verifiable, and cited.
Solution: Barry, Ontop’s CustomGPT.ai AI Agent
Tomas Giraldo evaluated multiple AI solutions before selecting CustomGPT.ai, citing ease of use and its business-grade capabilities as the deciding factors.
Using CustomGPT.ai’s no-code platform, Ontop built “Barry” an internal AI agent trained on company documentation covering legal requirements, payroll processes, EOR compliance, and jurisdiction-specific regulations.
What Barry does:
- Answers sales team questions drawn from Ontop’s internal documentation
- Provides citation-backed responses each answer references the exact source document
- Operates inside Slack, Ontop’s primary communication platform, with no additional login required
- Handles questions continuously without requiring legal team involvement
How it was deployed:
- Initial deployment connected Barry to Slack via a Zapier integration
- Ontop later migrated to CustomGPT.ai’s native Slack integration, reducing query latency further
- A dedicated Slack channel was created to centralize all Barry interactions for tracking
- A reporting dashboard was configured to monitor usage, acceptance rates, and identify the highest-volume question topics
- The knowledge base is refined on an ongoing basis as new documentation is added or updated
“Now we have a dedicated Slack channel, and we can recall and recapitulate everything and store it in a Google Sheet or any report to see who is asking more questions and in what areas.”
Tomas Giraldo, Product Manager, Ontop
How Slack Integration Improved Adoption
The decision to deploy Barry inside Slack rather than as a standalone AI portal was a key factor in driving rapid, sustained adoption.
Ontop’s sales team already lived in Slack. By putting Barry in their existing workflow, there was no new tool to learn, no separate login to remember, and no behavior change required. Sales reps could ask Barry a compliance question in the same place they were already messaging colleagues.
This matters because enterprise AI tools frequently fail not because the AI is poor, but because adoption is low. When the AI is embedded in the tool employees already use every day, the barrier to first use effectively disappears.
The dedicated Barry Slack channel also created transparency: legal team members could observe the questions being asked, verify that Barry’s answers were accurate, and identify where documentation needed strengthening — all without being pulled into individual conversations.
Before CustomGPT.ai vs. After CustomGPT.ai
| Factor | Before CustomGPT.ai | After CustomGPT.ai |
|---|---|---|
| Response time | 20 minutes per question | 20 seconds per question |
| Legal team hours on FAQs | 130+ hours per month | Redirected to strategic work |
| Questions requiring human response | 100+ per week | Near zero |
| Source citation on answers | None verbal or informal | Every answer includes a citation |
| Sales team autonomy | Low dependent on legal | High self-serve via Slack |
| Visibility into knowledge gaps | None | Dashboard + Slack channel tracking |
| Compliance accuracy risk | High informal delivery | Low documentation-grounded answers |
| Legal team focus | Reactive answering FAQs | Strategic high-value legal work |
Ontop’s Results
By deploying Barry through CustomGPT.ai, Ontop achieved measurable, sustained improvements across both its legal and sales teams:
- 130 hours saved per month legal team capacity freed from repetitive FAQ responses
- 60x faster response time from 20 minutes per question to 20 seconds
- 400+ complex questions answered monthly handled by Barry without human involvement
- 60% acceptance rate in a compliance-critical environment where accuracy is mandatory, this reflects strong trust in AI-generated, citation-backed answers
- Zero additional headcount efficiency gains achieved without expanding the legal or support team
- Real-time knowledge gap visibility Ontop can now see which topics generate the most questions, informing both sales training and documentation priorities
- Active exploration of new AI use cases Ontop identified additional deployment opportunities as internal adoption grew
Why It Worked
Slack-native access eliminated friction. Barry lives where Ontop’s sales team already works. No new tool, no new login, no change in behavior. This single decision drove adoption more than any other factor.
Citation-backed answers built trust in a high-stakes domain. In international payroll and employment law, an incorrect answer can have real consequences. Every Barry response references the specific source document used to generate it. Sales reps can verify answers before acting. Legal team members can trust that responses are grounded in approved documentation, not AI inference. This verifiability was not optional it was what made adoption possible.
Legal team relief was immediate and visible. With Barry handling 100+ questions per week, the legal team experienced measurable, concrete relief. 130 hours per month is an auditable number, one that justifies continued investment and expansion.
Analytics closed the feedback loop. CustomGPT.ai’s dashboard gave Ontop visibility into what questions were being asked, how often, and with what acceptance rate. This turned Barry from a static FAQ tool into a continuously improving knowledge system, surfacing training needs and documentation gaps that previously had no visibility.
No-code deployment kept iteration fast. CustomGPT.ai’s no-code platform meant Ontop could build, launch, and refine Barry without engineering resources. The migration from Zapier to native Slack integration was seamless and further improved query performance.
What Is an Internal AI Chatbot?
An internal AI chatbot is an AI assistant trained on a company’s own documentation policies, legal frameworks, product specifications, compliance guides that answers employee questions instantly, without requiring human expert involvement. Unlike public-facing chatbots, internal AI chatbots operate on private, proprietary data and are deployed within enterprise communication tools like Slack or Microsoft Teams.
At Ontop, Barry is an internal AI chatbot: trained exclusively on Ontop’s internal legal and payroll documentation, deployed inside Slack, and accessible only to Ontop’s employees. Its answers are grounded in real company knowledge, not general AI training data, and every response includes a citation to the source document.
Learn more about CustomGPT.ai’s enterprise AI chatbot capabilities and how it works.
How Did CustomGPT.ai Improve Ontop’s Sales Efficiency?
CustomGPT.ai improved Ontop’s sales efficiency by eliminating the information bottleneck between the sales and legal teams.
Before Barry, a sales rep with a compliance question had two options: wait 20 minutes for a legal response, or make an assumption and move forward. Neither was acceptable in a multi-jurisdiction EOR environment.
After Barry, the same question gets a cited answer in 20 seconds drawn from Ontop’s actual internal documentation, with the source document referenced so the rep can verify it. Sales reps spend less time waiting and more time selling. Deals move faster. Customer response times improve.
The 60% acceptance rate in a legally sensitive domain indicates that sales reps are not just using Barry, but trusting its answers enough to act on them.
Explore CustomGPT.ai for enterprise knowledge search and AI customer support.
Why Citation-Backed AI Answers Mattered for Ontop
A citation-backed AI answer is a response that includes a reference to the specific document used to generate it. This distinction was critical for Ontop.
Ontop operates in international employment law a domain where an inaccurate answer about payroll rules in a given country can create legal exposure. A confident-sounding but unverifiable AI answer would have been rejected by both the legal and sales teams.
By contrast, Barry’s citation-backed answers allow any recipient to check the original source document before acting. This made the AI trustworthy in a compliance context where trust is the only basis for adoption.
Citation-backed answers also create an audit trail. If a sales rep acts on a Barry response, there is a documented reference to the underlying policy. For regulated industries, this is not a convenience feature it is a compliance requirement.
CustomGPT.ai’s anti-hallucination accuracy technology and security and trust standards underpin every citation-backed response Barry delivers.
Lessons for Enterprise Teams
Put AI where people already work. Standalone AI portals have low adoption. Embedding Barry in Slack was the difference between a tool people try once and a tool people use every day.
Documentation quality determines AI quality. Barry’s answers are only as good as Ontop’s internal documents. The deployment made documentation hygiene a strategic priority, not a side project.
Measure acceptance, not just usage. A high query volume with low acceptance means the AI is not trusted. Ontop’s 60% acceptance rate in a legally sensitive domain is a meaningful signal of real-world utility.
Analytics reveal what you did not know you did not know. The dashboard tracking which questions Barry receives most frequently exposed knowledge gaps in Ontop’s sales training that no one had previously quantified.
Efficiency gains compound. 130 hours per month is 1,560 hours per year of legal team capacity. At the compensation rates of specialized legal professionals, that is a material cost avoidance achieved with a no-code SaaS tool deployed without engineering resources.
Conclusion
CustomGPT.ai gave Ontop a way to distribute specialized legal knowledge across the sales team at scale — without adding headcount, without increasing legal team workload, and without sacrificing the accuracy that a compliance-sensitive environment demands.
Barry, the CustomGPT.ai AI agent, saved Ontop’s legal team 130 hours per month, cut response times from 20 minutes to 20 seconds, and handled over 400 complex queries monthly. The 60% acceptance rate in a field where precision is mandatory reflects the practical trust that citation-backed, documentation-grounded AI answers can earn in enterprise environments.
For organizations facing the same internal knowledge distribution problem specialized expertise needed by many, held by few — the Ontop model is a replicable template.
Ready to build your own internal AI assistant?
Deploy a custom AI agent trained on your internal documentation. No code required. Get citation-backed answers your team can trust inside the tools they already use.
Try free for 7 days Talk to enterprise sales
Frequently Asked Questions
How did Ontop use CustomGPT.ai?
Ontop used CustomGPT.ai to build an internal AI agent called “Barry,” trained on company documentation covering legal requirements, payroll processes, and EOR compliance rules. Barry was integrated with Slack and deployed to answer repetitive compliance and payroll questions from the sales team — questions that had previously required direct involvement from the legal team.
What results did Ontop achieve with CustomGPT.ai?
Ontop’s legal team saved 130 hours per month. Response time dropped from 20 minutes to 20 seconds. Barry handles over 400 complex questions monthly and maintains a 60% acceptance rate in a legally sensitive, compliance-critical environment.
How much time did Ontop save with its AI chatbot?
Ontop saved 130 legal team hours per month by deploying Barry. That time was redirected from answering repetitive sales FAQs to strategic legal work, with no additional headcount required to achieve the gain.
How did Barry improve sales efficiency?
Barry gave Ontop’s sales team instant access to accurate, citation-backed answers to compliance and payroll questions. Response time dropped from 20 minutes to 20 seconds, allowing sales reps to get the information they needed without waiting for or interrupting the legal team. Sales could focus on selling instead of chasing answers.
Why was Slack integration important?
Slack is Ontop’s primary internal communication tool. Deploying Barry inside Slack removed all friction from adoption no new tool, no separate login, no behavior change required. Sales reps ask Barry questions in the same place they already send messages. This was a primary driver of sustained, high-volume daily usage.
What is an internal AI chatbot?
An internal AI chatbot is an AI assistant trained on a company’s own private documentation — policies, legal frameworks, compliance guides, product specs that answers employee questions instantly without requiring human expert involvement. It is deployed inside enterprise tools like Slack or Microsoft Teams and operates only on the organization’s proprietary data. Unlike public AI tools, it does not use general internet knowledge; its answers are grounded in the company’s actual internal content.
Can AI chatbots reduce legal team workload?
Yes. At Ontop, the CustomGPT.ai-powered AI chatbot handled over 100 questions per week that would otherwise have required legal team responses. This saved 130 hours per month and allowed legal professionals to redirect their time to strategic tasks. The enabling factor was citation-backed answers: because Barry referenced its source document in every response, the legal team trusted its accuracy and did not need to review or validate each answer individually.
What makes CustomGPT.ai useful for enterprise teams?
CustomGPT.ai is a no-code RAG (Retrieval-Augmented Generation) platform that lets enterprises build AI agents trained on their own documentation without engineering resources. Key capabilities include citation-backed answers that reference source documents, native Slack and API integrations, real-time usage analytics dashboards for tracking adoption and flagging knowledge gaps, and GDPR and SOC2 compliance for data security. Enterprises can deploy a production-ready internal AI agent in days, not months.
