Hey I'm Jesse, founder of Omnier!

I built a CRM that's way more token efficient and way faster compared to other legacy CRMs, when used by an agent.

The key figures are the average values from the tests below. All tasks were completed with Claude Opus 4.8 (on medium effort) with indentical CRM MCP setups. For full details: Test setup & full prompts

---

Create task

The first test in the benchmark is about creating new prospets (company + contact + deal + next-step task) from unstructured meeting notes. The full prompt includes 20 new prospects to be added into Attio/HubSpot/Omnier.

Prompt:

Hey add contact + company + deal for each of these new prospects in {CRM}. No need to check for duplicates, just simply create directly. No need to ask for permission, just create.

1. Met Jason Miller at SaaStr Annual, he's CTO at Flowbit (flowbitapp.com), closed a $8M Seed round in January led by Founders Fund. San Francisco, ~25 people. Building AI-native project management for engineering teams. Currently using Linear + Notion but looking to consolidate. jason.miller@flowbitapp.com, linkedin.com/in/jasonmiller. Wants to reconnect next month.
.
.
.

Result:

---

Search task

Second task is a search accross different deals using stage, contact linkedin, funding stage, and company size as filters.

Prompt:

List deals in prospect/qualified/demo stage where the main contact has linkedin url available, is C level / founder position, where the associated company last raised series A funding and has at least 10 employees, from {CRM}

Result:

---

Update task

The third task is a full deal context udpate after a succesful call.

Prompt:

The Parsio demo went really well — Ryan was impressed with the document-parsing accuracy. Update the whole deal context and move it to the negotiation stage, add a note summarizing that the demo covered their logistics use case and they asked about volume pricing, and create a task for me to send a pricing proposal by Friday. And also add Eric Johnson eric@parsioapp.com, linkedin.com/in/ericjohnson their CTOs info, he wants to be on the next call

Result:

---

Act + Report

The final task combines both search and update actions.

Prompt:

For every deal in Prospect with no activity in the last 14 days, create a high-priority follow-up task due Friday linked to the deal, and list the main contacts linkedin. Also give me total value of all deals that are going stale (no activity in 14 days) in my pipeline and the deal count per stage.

Result:

---

Evaluation

The greatest cost and speed difference emerges when tasks include more complex search queries, create jobs, or multistep actions.

For smaller changes, such as updating a single account after a meeting, the differnce to current CRMs is not as drastic. This can be seen from the Update Task result.

Omnier is built specifically for agents:

• The MCP tools expose direct SQL access -> agent can freely handle all CRM data.
• The database schema is built around json files, and is not scattered across dozens of different database tables like in legacy CRMs, which were originally built for human UIs.

The cost + speed gap between human CRMs and agent CRMs will increase as AI agents start to do more and more of the data handling completely autonomuously.

---

Try Omnier

Omnier lets you run ~3x more CRM work on the same Claude or Codex budget.

If you wanna test this yourself, here are the simple steps to connect: Getting started.

(Currently: unlimitied pilot credits for the first test users, limited time)

---

If you’re interested in learning more, send me a message at jesse@getomnier.com

Cheers, Jesse

---

Full Test Details

Key figure calculations:

cost multiplier(task) = platform cost(task) ÷ Omnier cost(task)
time multiplier(task) = platform time(task) ÷ Omnier time(task)

"3.9× cheaper" = mean(cost multiplier) across tasks with data on both sides

Raw test notes

• Attio RESULTS
• HubSpot RESULTS
• Omnier RESULTS

Test details

• Test setup & full prompts
• Full comparison