We fix things with AI.

We're two operators who build AI that does real work inside real businesses — the reporting, the customer conversations, the busywork that eats your team's week. Below are ten things we've actually shipped. If one of them looks like your problem, email us.

See the ten ↓ hello@fractionalfixers.ai
No pitch deck, no discovery-call gauntlet. Just work we can show you.
The work · 10 real cases

Things we've built that run every day

Each of these is live inside a real company right now. Client names withheld or shared on a call — the numbers are exact.

1Reporting

The morning scorecard that assembles itself

The problem

A wellness e-commerce brand started every day copying numbers out of six tabs — store, ads, refunds, bookings — and still couldn't see which number was right.

What we built

A nightly pipeline that pulls every source into one warehouse and publishes a single daily scorecard: sales, ad spend, refunds netted out, consults booked. Nobody assembles anything. It's just there at 7am.

16+ scattered reports became one screen, fed by 12+ live sources.

illustrative
Daily scorecard · auto-published 7:00a
$18.4kSales
212Units
$2.1kAd spend
2Customer-facing AI

A sales agent that actually knows the catalog

The problem

A B2B retailer's support team answered the same thirty compatibility questions all day — "will this work with my phone system?" — while buyers who couldn't wait simply left.

What we built

An AI agent living in the storefront that answers from the live product catalog — asking the questions a good salesperson would ask — and hands off cleanly to a human the moment it should.

Answers from real product data, measured after launch on deflection and what it got wrong.

illustrative
VisitorNeed wireless headsets that work with our Yealink phones — which ones?
AgentThree models fit Yealink out of the box. How many desks, and is the office open-plan? That changes my pick.
Visitor12 desks, open plan.
AgentThen noise-canceling mics matter — here's the one I'd start with, and I can loop in a human for volume pricing.
3Market intel

A price watcher across the whole market

The problem

An industrial distributor priced thousands of SKUs against competitors who changed prices silently. Checking by hand happened never.

What we built

A scheduled scraper that walks competitor sites on its own, with ten validation gates so a broken page can never poison the data, and screenshot evidence saved for every price it reports. The pricing team opens a dashboard, not forty tabs.

Every reported price carries its own evidence capture — trust built in, not assumed.

illustrative
Competitor sweep · last run 4:30a
7100 Series · rollrival −8%
Safety-Walk 24"we win
Bond 847 · caseno change
Grip tape · bulkrival −3%
Edge trim 50ftwe win
4Operations

A CRM rebuilt around how the team actually works

The problem

Orders lived in one system, fulfillment in another, and the team lived in spreadsheets between them. Nothing agreed with anything.

What we built

One CRM holding 10,000+ unified orders, with a ship-queue the warehouse runs from, address editing that syncs back to the carrier, and label printing built in. The spreadsheets retired themselves.

10,000+ orders unified · the whole team runs on one live screen.

illustrative
Ship queue · today
#21391 · 2 itemslabel printed
#21392 · 1 itemlabel printed
#21393 · 3 itemsaddress fix
#21394 · 1 itemqueued
5Finance

Month-end numbers the whole team actually reads

The problem

Every month the financials went out as a long, text-heavy email. Hours to write, skimmed in seconds, and no way to ask "why is that up?" without starting a reply-all thread.

What we built

The close now produces an interactive dashboard instead of a wall of text: revenue, margin, cash and spend against plan, with the trend behind every number one tap away. It assembles itself from the same figures as before; the finance lead reviews it and hits send.

Where the numbers live

Nothing new to log into and no outside service touching the books. It's a single self-contained file that travels inside the company's own email and access controls — exactly as private as the spreadsheet it replaced.

One of us wrote that email for years. Now the month lands as a dashboard, and the questions arrive pre-answered.

illustrative
Monthly financials · August close
OverviewP&LCashBy channel
+6.8%Revenue vs plan
41.2%Gross margin
+$14kOp-ex vs plan
6Paid media

One honest number for ad spend

The problem

The ad dashboards said things were great. The bank account disagreed. Every platform graded its own homework.

What we built

A profit-per-order floor computed from real margin — one ruler every campaign is measured against, with a day-one tripwire when a campaign dips under it. Spend decisions stopped being debates.

One profit floor replaced four competing "results" dashboards.

illustrative
Profit per order · by campaign
THE FLOOR
7Command center

The whole business on one screen

The problem

A dozen automated jobs, four companies' worth of work, and the owner's only view was "hope someone mentions it if something breaks."

What we built

A mission-control screen: per-company worklists, the status of every automated job, and a "needs you" feed that stays silent unless a human decision is actually required. Two check-ins a day, not forty.

Silent unless something needs a human — attention is the scarce resource.

illustrative
Needs you · 3 items
Approve refund over limit2m
Price change awaiting GO1h
Weekly numbers ready3h
All 14 jobs ran cleanauto
8Self-serve tools

A product editor the client team runs themselves

The problem

Every catalog change — a spec, a filter, a description across 3,000+ products — was a developer ticket with a two-week fuse.

What we built

An embedded editor inside their own store admin: search, stack filters, bulk-edit, apply. The team ships their own catalog changes now, and the dev queue lost its longest line.

3,000+ products, edited in-house, zero tickets.

illustrative
Bulk edit · 47 products selected
☑ 7100-BLK · anti-slip rollupdated
☑ 7100-YLW · anti-slip rollupdated
☑ 7150-GRY · tread 6"updated
☑ 44 more…applying
9Marketplaces

The Amazon console that replaced an agency

The problem

A brand paid an agency to manage Amazon and still couldn't answer basic questions about their own account, listings, or ad spend.

What we built

An eight-module operating console — sales, ads, listings, inventory, reclassification — wired to Amazon's own APIs. One in-house operator now runs the whole account, agency retired.

8 modules · one operator · one less agency invoice.

illustrative
Console · this week
+12%Amazon sales
4.2xAd ROAS
3Low stock
10Executive AI

An AI chief of staff for a CEO who will never open a terminal

The problem

A CEO wanted a personal AI assistant that just works — on his phone, every day — with no setup, no maintenance, and no patience for either.

What we built

A hardened, always-on assistant on dedicated infrastructure: hourly self-checks that re-apply their own fixes, and a one-message rollback for every feature. Running continuously since February 2026.

Six months of unbroken uptime for a user who never sees the machinery.

Meet the assistant →
illustrative
CEO · 6:12awhat's on deck today?
Assistant3 meetings, the board deck draft is ready for you, and your 2pm asked to move — want me to offer 4pm?
CEOyes. and remind me about the lease thing thursday
AssistantDone and done.
Who we are

Two operators, one team

We've each spent decades running the thing — not advising on it from outside. You get both of us; the problem decides who leads.

Chris BentleyCB

Chris Bentley

AI & Digital Operations

14+ years scaling e-commerce as a VP and P&L owner. Chris builds the systems on this page — the dashboards, the agents, the automations — and makes them boring enough to trust: monitored, verified, and owned after launch.

→ LinkedIn
Rick MillsRM

Rick Mills

Finance & Operations

20 years as an e-commerce CFO. Rick makes the numbers trustworthy — close cadence, forecasting, reporting owners actually rely on — and pairs it with the operating discipline that keeps growth from outrunning the business.

→ LinkedIn

"The rare combination is knowing what the technology can do and what the business actually needs. That overlap is where we live."

How it works

Simple on purpose

1.

Tell us what hurts

One email. Describe the problem in the words you'd use complaining about it — that's usually the most accurate spec we get.

2.

We show you the fix

A short call, then a concrete proposal: what we'd build, what it changes, what it costs. Often we've built something close before — you get that head start.

3.

You own what we build

One monthly number, published on this page before you ever talk to us. Every system we build runs in your accounts, on your infrastructure, and stays yours.

Pricing

Two numbers, both here

No sitting through a call to hear the price. Here it is.

$2,800per month
15 to 20 hours per month of solution execution reserved for you.

One system at a time, built and kept running.

or
$5,000per month
30 to 40 hours per month of solution execution reserved for you.

Several threads at once, or one build you want finished sooner.

Need more hours?

Additional hours are $160/hr.

  • No setup fee and no onboarding fee
  • Everything runs in your system or on ours, whichever you prefer. The code is yours.
  • Monitored after launch, not handed over and forgotten
  • Month to month, no contract

If the work doesn't fit a monthly rhythm, we'll work to price it as a project instead.

Questions people ask first

The things you'd ask on the call

Answered here so you don't have to book a call to find out.

What does "fractional" mean?

You get an experienced operator part time, without the full-time hire. You buy the specific work you need instead of a salary, a headcount slot and a hiring cycle. We do it for AI systems, operations and finance.

What do you actually build?

Working solutions that runs on its own inside your business. Daily reporting that assembles itself, AI agents that answer customer questions from your real product data, price monitoring, order systems, internal tools your team can operate without a developer.

How long does it take?

It depends entirely on what's broken. Some of the systems on this page were useful within hours; others took months of iteration to get right. We'll tell you which one yours looks like on the first call, before you commit to anything.

Do we own what you build?

Yes. It runs in your accounts, on your infrastructure, under your logins. If you stop working with us it keeps running, and the code is yours.

What happens if the AI gets something wrong?

We assume it will, sometimes, and we build for that. Validation gates on incoming data, evidence captured for anything the system claims, a human handoff on anything that needs judgment, and a rollback path for every feature. We give you an AI system you can audit at any time.

Who is this for?

Owner-run and mid-sized companies where one broken process quietly costs real money every week. Less useful if you already have an internal platform team, or if you are pre-revenue and really looking for a technical cofounder.

What happens on the first call?

About thirty minutes. You describe what you wanted fixed or improved. We tell you roughly what it would take, and if we are the right or wrong people for it. No deck, no discovery sequence, no second call to qualify you.

Got something that needs fixing?

Tell us what's broken, slow, or eating your team's week. If we can help, we'll say how. If we can't, we'll say that too.

One inbox · both of us read it · replies from a human