Our work · Names withheld

Things we built that sales teams use every day.

Three solutions you can see on a call, plus six older engagements behind a shared password.

Public preview only
02 · Solutions

Three things we built that sales teams use every day.

We keep the details for the call. Each one runs inside a client's stack today. Here is what changes for the reps.

Live with clients
On the phone · after every call

Call Notes Agent

A rep records the call on the phone they already carry. Minutes later the deal has a transcript, a summary and the next steps. Nobody typed a word.

  • Notes land on the right deal, not in a notebook
  • Managers read the summary, not the recording
On the call, you'll see how it decides what to keep.
Live with clients
Inside Outlook · all day

Sales Inbox for Outlook

A pane inside Outlook finds the emails a rep must answer today. It drafts the reply and files the thread to the right deal. The CRM stays closed.

  • Buyers get a reply the same morning
  • Every thread logged without anyone logging it
On the call, you'll see how it picks the emails that matter.
Private client
Inside the CRM · every deal

AI Sales Closer

An AI closer plugged into your CRM through a connection your IT team controls. It watches every open deal and drafts the next move. It prices within the rules and asks a human before anything is sent.

  • Quiet deals get a nudge before they die
  • Every action logged, nothing sent unseen
Built for a client we can't name. We show it with their permission.
  1. 1
    Free pipeline review 30 minutes. Where do deals get stuck.
  2. 2
    Walkthrough of the one that fits Live, on your data if you want.
  3. 3
    Fixed-price package Quoted after the review. No per-seat fee.
Case studies

Older engagements, in detail.

Security, data, and custom builds for IT and security teams. Unlock to read the challenge, approach, and outcome.

Security Wealth Management

Salesforce access review

A 1,200-user Salesforce system with eight years of piled-up permissions. We mapped who could see what, and what outside apps could reach.

78 Unused access bundles removed
Challenge

Eight years of growth left 234 access bundles — Salesforce calls them permission sets. 78 sat unused, assigned to nobody, but still active. One outside app could read 14 fields it had no reason to see. In one case, sharing settings let wealth data cross a line it should not have.

Approach

A two-week review. We listed every access bundle and who held it. We checked what each connected app could reach. We traced the sharing settings for leaks and ranked every finding by severity and impact. An engineer was paired with each serious item to fix it.

Outcome

78 unused access bundles removed. Two connected apps limited to what they need. One sharing setting rewritten to close the leak. The attack-surface score dropped 41%. Nothing users relied on broke.

SalesforceApp access reviewSharing rulesSeverity scoring
Agentic AI Payments

AI helper boundary review

Five AI helpers, four connection points, twelve tools. We mapped what each helper could actually reach against what the policy said.

4.2x Faster reviewer queue
Challenge

A fast-moving fintech had five AI helpers sharing one connection with wide access. The fraud helper could read commission tables it had no business seeing. The scheduling helper could write through a tool the team had forgotten about. Nothing was logged for replay.

Approach

A one-week review. We listed every tool each helper could use. We recommended tight limits per helper and tried to trick the helpers with hostile prompts. Then we replayed 30 days of their activity. Finally, we rewrote the setup so each helper gets only the tools its job needs.

Outcome

Each of the five helpers now has its own limited set of tools. The reviewer queue shrank because structured answers cut false alarms. Throughput rose 4.2x. Auditors signed off by reading the schemas.

AI helpersMCP (AI-to-data connections)Hostile-prompt testingReplay logs
Custom Build Healthcare Network

Security monitoring with audit reports built in

A monitoring system built around the client's own stack. It produces SOC 2, HIPAA, and GDPR evidence on every release.

11→2 Compliance analysts freed
Challenge

A regional healthcare network was spending eleven analysts a quarter assembling SOC 2 and HIPAA evidence. Their off-the-shelf tool produced reports the auditors rejected, so everything was rebuilt by hand from raw logs in spreadsheets.

Approach

A 90-day build. We designed the system around their cloud and on-site data, and tuned alerts to their HIPAA risks. We made control checks run on every release. We built dashboards for the security team and leadership, then handed it over with our engineers alongside theirs.

Outcome

Manual compliance work dropped from 11 analysts to 2, who now review automated output instead of compiling it. Control checks run on every release. Audit prep went from 6 weeks to 3 days.

Custom monitoringHIPAASOC 2Cloud and on-site data
Data Fintech

Data warehouse access cleanup

A four-year-old Snowflake account with 38 roles and 12 service accounts. Three reporting tools could see more than they should.

−63% Shorter path to sensitive data
Challenge

Roles had blurred together over time. The reporting service account could read tables it never used. Default warehouses were open to everyone. Everything worked, which was the problem: nobody had a reason to look.

Approach

A two-week review. We walked the role structure from the top down and documented every grant and why it existed. We searched 90 days of query history for unusual activity, then traced where sensitive data flowed into reports.

Outcome

14 role inheritances flattened, 23 unjustified grants removed, and three reporting accounts limited to the datasets they actually show. The path from a public role to sensitive tables got 63% shorter.

SnowflakeSigmaAccess reviewData flow tracing
Custom Build Logistics

Field-service app that works offline

A custom app where field reps log site visits without a signal. It syncs to the books for invoicing when they reconnect.

2.1d Days cut from each invoice
Challenge

Field reps filled in paper forms and photographed them. They emailed the photos to the office, where staff re-typed everything into Zoho Books. Errors piled up. Invoices took four to six days after the work was done.

Approach

A four-week build. It started with one planning workshop with two field reps and the bookkeeper. Then three weeks of engineering built a Zoho Creator app with forms that work offline. A bridge into Zoho Books catches duplicates and routes approvals. A security review ran before launch.

Outcome

Invoices now go out one to two days after the work, down from four to six. 89% of field reps were using it in the first week. No data lost in six months of use.

Zoho CreatorZoho BooksOffline formsApproval routing
Agentic AI Fintech

Fraud review helper with a full audit trail

An AI helper that reads a transaction and checks four data sources. It returns a fraud verdict it can justify line by line.

0.4s Median time to a verdict
Challenge

A fintech's fraud reviewers were overwhelmed. The old rules engine missed new patterns, and a chatbot trial could not explain its decisions. They needed something an auditor could trust.

Approach

We built an AI helper with controlled access to four sources. It can reach transaction history, identity checks, sanctions lists, and device data. Every lookup is recorded and can be replayed. Every verdict cites the exact data it used.

Outcome

The reviewer queue fell 71%. The helper clears 73% of transactions on its own and escalates 27% with a written reason. A verdict takes 0.4 seconds. Auditors approved it because they could replay any decision.

AI helperControlled data accessReplayable decisionsAudit trail
Take the next step

Give your reps their selling hours back.

Thirty minutes on a call. You leave with a list of what comes off your reps' plates first, and a fixed quote if you want one. No deck.