Gumloop | Build, optimize, and control AI agents

Build AI agents for work

Teams get AI agents. IT gets security and oversight. Leadership sees the impact.

Roll out specialized agents in minutes

Data Analysis Agent

A reasoning agent that answers questions from your data warehouse.

Data - Built by Aron, Katherine, Max

Support Agent

An agent that triages bugs, creates tickets, and spots support patterns automatically.

Support - Built by Gonzalo, Katherine, Marcelo

CRM Agent

An agent that manages deals, researches prospects, and keeps your CRM up to date.

Sales - Built by Aron, Aron, Rahul

Meeting Prep Agent

An agent that automatically briefs you before every meeting with context from across your tools.

Sales - Built by Gonzalo, Max, Rahul

Call Analysis Agent

An agent that analyzes call recordings to surface objection patterns, coaching insights, and competitive intelligence.

Sales - Built by Aron, Katherine, Marcelo

Where are we losing people in the onboarding flow?

Gumloop

4 steps

I should query BigQuery for onboarding funnel data to identify where users are dropping off. Let me explore the available event types first to reconstruct a funnel.

Querying available event types

Good — I have enough signal to reconstruct an onboarding funnel using login → dashboard → workflow_created → workflow_run as a proxy for the key steps. Let me build the analysis.

Building onboarding funnel query

The biggest drop-off is between Viewed Dashboard and Attempted Integration at 46%. This is the #1 problem area. Let me generate a visualization.

Building onboarding funnel chart

Here's where you're losing people:

🚨 Biggest drop-off: Dashboard → Attempted Integration 46% of users who view the dashboard never even try to connect an integration — that's 1,432 people falling off in a single step. This is your #1 problem.

⚠️ Secondary drop-off: Attempted → Completed Integration Of the users who do try, 37% fail to complete it. Combined, only 22% of signups make it through the integration step at all.

Step Users Drop-off
Signed Up 4,820 —
Completed Profile 3,940 18%
Viewed Dashboard 3,105 21%
Attempted Integration 1,673 ⚠️ 46%
Completed Integration 1,056 ⚠️ 37%
Ran First Automation 712 33%
Fully Onboarded 481 32%

Only 10% of signups make it all the way through onboarding. The integration step is clearly the wall. Want me to break this down by acquisition channel, plan tier, or look at which specific integrations are failing most?

Does it depend where they came from?

Gumloop

3 steps

The user wants to know if the onboarding drop-off depends on acquisition channel. Let me query channel-specific funnel data and create a comparison.

Checking customers table schema

Querying channel funnel breakdown

Building channel breakdown chart

Yes — channel makes a massive difference. Here's the breakdown:

🟢 Referral is your best performer by far 38% of referral signups fully onboard. They're the most motivated to connect an integration (70% attempt it) and complete it at the highest rate (87%).

🟡 Organic is strong on completion, weak on attempt Organic users who try an integration almost always finish (99% completion rate) — but only 48% ever attempt one.

🔴 Paid Ads is a black hole 1,540 signups — your biggest volume channel — but only 4.7% fully onboard. Only 40% attempt an integration, and of those, barely half finish.

Channel Signups Attempted Completed Onboarded Conv %
Referral 710 70% 87% 270 38%
Organic Search 1,820 48% 99% 456 25%
Direct 510 50% 80% 98 19%
Product Hunt / Social 240 36% 72% 22 9%
Paid Ads 1,540 40% 47% 72 ⚠️ 5%

The two clearest actions:

  1. Investigate why paid users abandon the integration step — are they the wrong ICP, or is the onboarding experience failing them specifically?
  2. Invest more in referral — it converts at 8× the rate of paid. Even a modest referral incentive program could move the needle significantly.

Weekly Active Users

W1

W6

W12

W18

W24

Anyone builds, everyone benefits

Gumloop is powerful enough to run the world's largest companies, but simple enough that anyone on any team can start building AI agents in minutes. No code, no additional subscriptions, or add-ons required.

Empower your experts to build

Technical and non-technical teams alike can collaborate on AI agents together, and deploy them where the team already works.

Connect and consolidate your tools

Choose your favorite models and apps from Gumloop's hundreds of integrations.

Coordinate teams of agents

Orchestrate complex processes with multiple agents inside deterministic workflows. Agents can even call on subagents.

Agents that run themselves

Tell Gumloop in plain English what to watch for on the web or across any combination of apps, and it'll monitor and take action automatically.

Automatically write, execute, and refine skills

Agents can create, edit, and improve their own skills at runtime, so they get better and more efficient.

Create live dashboards, decks, and apps

With just a prompt, build shareable files and web apps that pull live data from your connected tools.

Welcome agents into your workplace

Interact with agents like co-workers in Slack, Teams, and email. Simply tag @Gumloop and ask.

See the spend, cut the waste

Know exactly where your compute and cost is going, for every agent and every team. Get the most value out of every credit and proactively avoid runaway costs, instead of waiting for the bill.

Spend, attributed

Every tool call and every run is tied to an agent, a team, and an owner. No more black-box invoices.

Swap models freely, with no vendor lock-in

Move any agent to a more cost-effective or higher-performing model.

Full visibility, full control

Your data stays yours. No shadow AI on team laptops, no data leaking to model providers, no agents running unobserved. Agents run where you want, how you want, only in the ways you've approved.

Role-based access controls

Manage reusable roles with granular scope control and custom permissions.

Secure credential management

Centralize secret storage and manage shared credentials with scoped access controls.

Audit logging

Get complete visibility with detailed audit trails for actions across the org.

Reporting and analytics

Monitor AI adoption, usage, and outcomes across all teams in one place.

Usage monitoring

Track org-wide credit usage in real time. Implement budget and quota controls to avoid surprises.

AI model restrictions

Control which AI models teams can (or can't) use, and set global fallback models. Implement guardrails and spend policies.

“The partnership review agent cuts prep time from 30–45 minutes per customer to under five minutes, and it results in a 31% higher win rate and 44% larger deals.” Matt GouldHead of Insights and Operations, Gusto

Latest updates

Meet Gumball (Beta)

Gumball is your personal agent. It writes you a morning newsletter called the Daily Chew, briefs you before each meeting with Meeting Prep, and sorts your email with Smart Inbox, drafting replies and flagging what needs your attention. Open it from your sidebar, chat in Slack, or email gumball@gumloopagents.com.

Custom Links for Artifacts

You can now publish an artifact on its own URL, like sales-dashboard.gumloopartifacts.com. Turn on Custom Domain from the share dialog, pick a name, and hand out a link people can actually remember. Access and permissions stay exactly the same.

FAQs

What's a Gumloop agent?

Do I need to know how to code?

Which AI models can I use?

How do agents connect to my tools?

Where do my agents run, and is my data safe?

How do I control and audit what agents are doing?

Can I see what agents are costing me?

Can agents work together on bigger processes?

Do agents get better over time?

How does my team use the agents day-to-day?

Build your first agent