Surphmore
AI consulting + custom builds

WebuildtheAIsoftwareyourbusinessactuallyneeds.

Most AI projects fail because nobody scoped them. We scope them, build them, and ship them in six weeks. No demos that fall apart in production.

05Who you're hiring

You're not hiring an agency.

Marc Surphlis, founder of Surphmore Consulting
Marc SurphlisFounder

Hi, I'm Marc. Surphmore is a consultancy of one. The work doesn't get pushed to interns or offshore subcontractors. I scope it, I write it, I ship it.

Twelve years building software. The last three running real AI in production for clients who needed it to actually work, not just demo well. A Stripe-grade engineering shop in a kayak.

What I won't do:

  • Train models on your data without telling you, ever.
  • Pretend a wrapper around someone else's model is a product.
  • Take on a project I'm not sure I can ship.
0+
years shipping
0 yr
AI in production
0
abandoned projects
02How we work

Six weeks, four moves.

01
Week 0 · Scope

Two sessions. One real plan.

We map what is hurting, what would actually move the number, and what is not worth touching yet. You leave with a written scope and a fixed price.

02
Weeks 1–2 · Prototype

Working version inside week two.

You touch it. We watch what you do with it. Then we cut what nobody used and double down on what landed. No guessing, no wireframes living in a Figma file.

03
Weeks 3–6 · Ship

Live for your team by week six.

Production code, your stack, your repo, your cloud. No black box, no rented platform, no vendor lock-in. We hand you the keys and walk you through them.

04
Week 6+ · Scale

Stay on retainer or wave goodbye.

We stick around if you want a second build or ongoing tuning. Most clients ask. Some hand off to their team and we disappear. Both are fine outcomes.

Worth a thirty-minute call?

Bring your problem.
Leave with a plan.

03Recent work

Built for operators.

Three from the last twelve months. The clients are real. The metrics are theirs. Names redacted because most of these are still our clients.

Dispatcher
Got a load Charlotte → Atlanta?
Agent
Booked DAT #82461. Pickup 6am. Risk: low.
Agent
Customer pinged. ETA at 11:42.
01Logistics2026

Fleet dispatch agent

40-truck regional carrier
Problem

Dispatcher spent six hours a shift in spreadsheets, juggling load boards and customer updates. Mistakes were expensive and tired humans were making them.

What we built

An AI agent that books loads off DAT, flags risk against historical lanes, and pushes proactive updates to the customer. Plugs into their TMS as just another user.

Outcome

Two hours back per dispatcher per shift. Late-pickup rate down by a third. The dispatcher no longer threatens to quit on Mondays.

2 hrssaved per shift
triage queue
92Acme Corp · 250 seatsHOT
71Initech · 80 seatsWARM
38Hooli Labs · trialNURTURE
12test@gmail.comFILTER
02B2B SaaS2026

Inbound triage + demo book

$4M ARR vertical software co.
Problem

Enterprise inquiries waited 14 hours for a first reply. Half went cold. Sales blamed marketing. Marketing blamed sales. Both were right.

What we built

Triage agent that classifies, enriches, and routes inbound. Hot leads get a Calendly link before they close the tab. Everyone else gets a real human draft for review.

Outcome

First-response time down to under ten minutes. Demo book rate doubled in one quarter. Two SDRs reassigned to outbound that actually moves the needle.

9 minfirst response
Equipment ledger · Wk 18
Live
$48,920
178 assets active5 branches
Br. 12
82%
Br. 22
71%
Br. 35
94%
Br. 41
58%
Br. 88
77%
A-2014Br. 3540 hr$1,840
B-1872Br. 2232 hr$1,420
C-2008Br. 12OFFcredit
03Industrial rental2026

Full-scale inventory and invoicing platform

Multi-division operator, 5 regional branches, 180+ assets
Problem

Five interlinked spreadsheets ran an eight-figure equipment business. Every rate change, off-rent, or transfer meant re-keying the same data into five files. One operator held all the formula knowledge in their head, and rate errors quietly leaked into customer invoices.

What we built

A full-scale corporate inventory and invoicing platform. Captures equipment moves once and routes them through approval, scheduling, and billing in one flow. Retroactive editing rebuilds closed months as credit notes instead of voided invoices. Role-based access for ops, admin, and leadership. Two years of historical data was migrated and reconciled line by line against the source spreadsheets.

Outcome

Now handling $2.1M of rental revenue across 178 assets and 3,300+ rent records with zero formula-related billing discrepancies. Weekly invoice generation went from four hours to under ten minutes. Eliminated the dependency on a single operator and recovered roughly ten hours of admin time per week.

$72kannual savings
$2.1Mrental revenue managed
4hr→10mweekly invoicing
01What we build

Five flavors of work. All of them ship.

We do not do whitepapers. We do not do six-month strategy decks. We pick the one or two things that move your number, and we write the code.

i.

AI agents

Custom agents that handle a real job inside your business. Sales follow-ups. Inbound triage. Internal research. They use your data, your tools, your tone.

  • Tool-using agents on Anthropic + OpenAI
  • Stateful, evaluable, observable
  • Owned by your team, not us
ii.

Workflow automation

We pick what's slow and what humans hate, then we make it disappear. Connects to whatever software you already pay for.

  • Zapier-grade reliability, custom-grade logic
  • Bidirectional sync to your stack
  • Audit trails for everything
iii.

Custom internal tools

The dashboard your team has been begging the Notion page to be. Real database, real auth, real multi-user. Built in weeks, not quarters.

  • Production Next.js + Postgres
  • Role-based access from day one
  • Yours forever, your repo
iv.

Fractional AI engineer

Drop me into your team for a quarter. I act as your senior AI engineer in your standups, your repo, and your Slack, shipping the highest-leverage things first. No middle layer, no PM hand-off, no junior who needs hand-holding. The decisions get made in the same head that writes the code.

  • One senior engineer, no team to coordinate
  • Working in your Slack, your standups, your repo
  • Weekly demos, clean handoff at the end
v.

Front-end websites

Marketing sites and product UIs that look like your engineering team is bigger than it is. Built fast, animated tastefully, accessible by default. This site is one of them.

  • Next.js + Tailwind + Framer Motion
  • Lighthouse 95+ on mobile, every page
  • Yours in your repo, deploy where you want
04Built with

Tools we work in.

Open-source and well-maintained. What your team can actually own after we leave.

OpenAI
Anthropic
Vercel
Next.js
Supabase
Postgres
Stripe
Linear
Notion
Slack
TypeScript
Python
Let's build

Tell us what's
slow.

Walk away from a thirty-minute call with a real plan. No pitch, no slide deck, no obligation. We will tell you if we are not the right fit.

Marc Surphlis

I'll be the one on the call. No SDR, no junior, just me. Bring your problem and we'll spend the first ten minutes on what's hurting, fifteen sketching what to build, and five on whether it's worth doing.

— Marc Surphlis
cal.com / surphmore / discoverycall
Calendar opens in a new tab
Pick a thirty-minute slot that works. We will be on the call.
Open the calendar