Business Technologist Yunpeng Wang · New York

A business technologist’s practice

I run autonomous AI agents in production. I’ll build ones for your business.

In my day job I’m a Principal Data Analyst at WeightWatchers, where I build and run autonomous AI systems that handle real operations — real money, real deadlines, real consequences if they break. I do this 12 hours a day.

This is my personal practice. I work with one or two companies at a time to build the same kind of systems for them.

Who this is for

Operations that outgrew spreadsheets and Zapier.

You run or operate a 20-to-30-person company. Your team is competent. But your operations have outgrown spreadsheets and Zapier, and someone — often you — is spending hours every week on work that doesn’t require human judgment:

  • Reconciling numbers across tools that should already agree
  • Reading dashboards looking for what changed
  • Chasing commitments people made and didn’t track
  • Reviewing routine work before it ships
  • Manually posting, responding, categorizing, tagging

You’ve tried hiring an ops person, an analyst, a VA. Each one helped, but the work scaled faster than the headcount. You don’t want to build a 50-person ops team to support a 30-person business.

What I actually build

Not chatbots. Autonomous systems.

They wake up on a schedule, pull data from your tools, make decisions inside guardrails you define, verify their own work, and either act or surface what needs your attention.

The operational morning brief

Every morning at 6 AM, a system reads your CRM, accounting, project management, and inbox since yesterday and writes you a one-page summary: yesterday’s numbers vs. plan, the three things that need your attention today, commitments people made and haven’t followed through on, prep for your calendar. Not a dashboard. A written brief, waiting when you wake up.

Budget and spend guardrails

Agents monitor your ad accounts, inventory orders, payroll runs, or cloud bills and flag anomalies within minutes — not at month-end when the damage is done.

Review and response automation

Customer reviews get summarized weekly with the patterns highlighted. Responses get drafted in your voice for approval. Negative reviews trigger a Slack notification before they sit unanswered for three days.

Customer service triage

Inbound emails get classified, prioritized, and drafted-response-attached before a human looks at them. Routine ones auto-send with a safety check; complex ones route to the right person with full context.

Social and content posting

A system that drafts, schedules, and posts to your channels in your brand voice — with a 30-second human approval step instead of a 30-minute one.

Financial close acceleration

Reconciliation work that takes a bookkeeper days gets reduced to hours, with AI doing the matching and a human confirming the edge cases.

The pattern across all of these is the same: a recurring loop that today consumes human attention, replaced by a system that handles 80% autonomously and routes the rest to a human with full context.

The verification layer

Nothing that touches real money runs unchecked.

Anything that touches real money has a second AI cross-checking the first against your source of truth. No system makes a decision on a hallucinated number. No one gets fired because an AI got creative.

This is the part most AI consultants skip. It’s also the part that determines whether you can actually trust the system enough to step away from it.

Why me

I do this in production, every day.

Most people selling AI consulting watched a YouTube video. I run these systems in production at a Fortune 500 consumer brand, in an environment with real compliance constraints, real auditors, and real budget consequences when something goes wrong.

I know which loops are worth automating, which ones look automatable but aren’t, where the failure modes hide, and how to build systems that survive the week you’re on vacation.

The hard part of this work isn’t the AI. It’s understanding operations well enough to know which automations earn their keep and which ones become technical debt. That’s what I bring.

How working together looks

A working system in production by the end of the month.

WEEK 1

Discovery sprint

Two or three working sessions where I learn your actual operations. Not your “AI strategy” — your Tuesday morning. You leave with a short list of automation candidates ranked by ROI. Fixed fee, credited toward the build if you proceed.

WEEKS 2–5

Build the first system

We pick one loop. I build it. You test it. We iterate. By the end of the month you have a working system in production.

ONGOING

Expand or step back

If the first system earns its keep, we build more. If not, we don’t. Everything I build is documented and runs on your infrastructure. You’re never locked in.

Engagement and pricing

Straightforward terms.

Discovery sprint $2,000
Initial engagement from $10,000
Typical timeline 4–6 weeks
Optional retainer $2,000–$4,000 / mo
Concurrent clients two at a time

The discovery sprint is credited in full toward the engagement if you proceed. Retainer covers monitoring, tweaks, and additional builds. Everything documented and running on your infrastructure.

Starting a conversation

If you can describe in one sentence the loop in your operations that’s eating the most time — let’s talk. The first 30 minutes are on me.

Book a free 30-minute call

Prefer email? yunpeng.wang@businesstechnologist.ai