Engagements told plainly. Client names appear only with permission.
Wealth management, $500M to $1B in assets · Governed AI and policy alignment
AI for advisors, inside the firm's own Microsoft 365, with controls the compliance team could document.
The problem
Advisors wanted to use AI with client information. Leadership and the firm's outside compliance consultants needed to know exactly what controls would exist before the AI policy in the compliance manual could be finalized.
What we did
Built a governed AI assistant inside the firm's own Microsoft 365 tenant. Staff sign in as themselves; answers come from approved sources; lookups into portfolio reporting are read-only; every interaction is retained under the firm's policy, searchable in eDiscovery and covered by data loss prevention. Administration runs on least-privilege access with dedicated break-glass accounts. We then wrote up how the assistant works as built and its control environment, and mapped it section by section to the firm's draft AI policy, with suggested changes for the firm to accept or leave. The policy itself stays with the firm.
The result
Advisors are onboarded and using it, and the firm's outside compliance consultants are using our control write-up to finalize the policy. Live for advisors within three weeks of signing.
"Thanks for sending over the breakdown of [the assistant] and the control environment. We think it serves well to capture controls and usage of the system and its features."
The firm's outside compliance consultant
Microsoft 365
Copilot Studio
Claude
Microsoft Purview
Entra ID
Black Diamond
References available on request.
How we approach it. Examples are illustrative.
Flooring contractors · Operations engine
RPE Systems: an operations engine that blocks bad jobs before they start
The problem
Sales lived in the CRM, jobs lived in the project tool, and margin lived in someone's head. Owners found out a job was unprofitable after the crew was paid.
What we did
Built the engine behind RPE Systems. When a deal is sold, the system checks margin against the owner's rule (40% gross margin floor, 15% labor cap) and holds anything below it for approval. Jobs, change orders and labor flow between GoHighLevel and ClickUp through n8n. Owners, GMs and controllers each get their own live dashboard view. The workflows were built as templates, so a second flooring contractor went live on the same engine.
The result
One pipeline from new lead to closed job, with alerts before problems turn into losses, documented with SOPs and a turnover pack.
"What stood out most was his ability to troubleshoot difficult issues and stay committed until the solution was fully working."
Carey Hunter, RPE Systems
GoHighLevel
n8n
ClickUp
Google Sheets
Slack
Online membership business · AI assistant
An AI concierge that remembers every conversation
The problem
Leads arrived on WhatsApp and Instagram at all hours. A generic chatbot forgot context and pushed junk data into the CRM.
What we did
Designed and built an AI concierge with real conversation memory, intent handling and clean CRM records, in phases, then a website version for free members.
The result
The client came back for two more builds.
"He clearly understands how to design scalable systems using GPT, ManyChat, n8n, and Airtable, and more importantly, why each piece should exist rather than over-engineering prematurely."
AI concierge client, Upwork
OpenAI
ManyChat
n8n
Airtable
WhatsApp
Instagram
Apparel production · Systems review and fix
CRM and ERP review: why the sales board was lying
The problem
Cards sat in the wrong pipeline lanes and deal values kept changing. The integration between the CRM, the automation layer and a custom ERP had been edited by several people and nobody could say why.
What we did
A read-only review first. We found two workflows taking turns overwriting the same deal value, a live workflow edited outside version control that had lost its retry logic, and a default ERP list that silently dropped expired quotes, so one business rule could never fire. We wrote the fix offline and tested it against real captured records before touching anything live.
The result
The patched logic made the same stage decisions as production on 19 of 19 real records while fixing titles and values, and a dry run showed exactly which deals would be renamed before anything ran. The client moved straight into an implementation phase.
"Working with Richard has been great. He's very thorough."
Apparel production client, Upwork
GoHighLevel
n8n
Custom Rails ERP
Wealth management · Automation diagnostic
An inherited AI content pipeline, diagnosed without touching production
The problem
A content pipeline built by someone else: five Make.com scenarios, 338 modules, eight validation gates and four AI calls, with a platform deprecation deadline approaching and nobody sure what it really did.
What we did
Cloned everything into a sandbox and audited it there. Mapped every scenario and gate, flagged the deadline risk across more than 100 modules, and found a quality gate hardcoded to a test value that was passing every post. Delivered a written system reference and a recorded walkthrough the COO could share with the firm's compliance lead, plus a phased fix plan.
The result
The firm knew exactly what was running, what was broken and what to fix first, with no production change during the diagnostic.
"You will not find a Make.com specialist who is (1) clearer at communication, or (2) of a higher caliber in terms of nuanced understanding of being truly solutions-oriented. Period!"
We work with wealth management and other regulated firms. We build the technology and keep the documentation ready; the firm keeps its compliance decisions.
Graphics in videos marked illustrative are examples, not client data.
Tell us what is not working.
A system that should talk to another one and does not. An automation nobody can explain. A technology decision you would rather not make alone. Book 30 minutes and you will leave with a straight answer and a next step.