Customers

Real businesses. Real workflows. Real ROI.

Each story below is a Forward Deployed Engineering deployment with measured outcomes. We don't share customer names without explicit permission — but the workflows and numbers are real.

Case 01·Logistics · Freight Broker

From 30-minute manual quotes to 30-second AI quotes.

Challenge

Brokers were spending 30 minutes per LTL quote — checking DAT, calling 3 carriers, calculating margins. Capacity capped at 16 quotes per quoter per day. Manual error rate ±15%.

Workflow we shipped
  1. ›Receive — Web form / email / WeChat captured into one queue
  2. ›Analyze — DAT API + historical pricing model + risk classifier
  3. ›Notify — Quote draft + margin signal pushed to quoter in 30s
  4. ›Assign — Approved quotes auto-routed to carrier matching

$280K+ annual labor savings.

Next step: Now packaged as an L3-P3 Enterprise Deployment offering for other freight brokers.

Outcomes
Time per quote
30 min→30 sec
Quotes per quoter / day
16→200+
Pricing error
±15%→±3%
Headcount for same volume
10→2
Case 02·Restaurant · Flushing

Reply to every Yelp review in 15 minutes — bilingual, in-voice, on-brand.

Challenge

200+ Yelp / Google reviews per month. Owner spent 30 hours/month replying. Star rating drifted 4.5 → 4.1 because negative reviews went unanswered.

Workflow we shipped
  1. ›Receive — Yelp + Google APIs polled every 15 minutes
  2. ›Analyze — Sentiment + dish mentioned + service category
  3. ›Notify — Positive replies auto-sent; negatives drafted + WeChat ping to owner
  4. ›Assign — Weekly trend report: dishes praised / hours complained about

Tooling cost: $50/month. ROI in week 2.

Next step: Owner enrolled in L2-V1 (Restaurant AI) to internalize the playbook for menu + inventory.

Outcomes
Owner time / month
30 hours→2 hours
Reply time
24 hours→15 min
Yelp star rating
4.1→4.5
Monthly revenue
Baseline→+12%
Case 03·Accounting · CPA Firm

200 → 500+ tax-season clients without adding a single bookkeeper.

Challenge

3 bookkeepers entered 500 receipts/day each at 8% error rate. Owner stayed up to 2 AM reconciling. Tax season capped client capacity at 200.

Workflow we shipped
  1. ›Receive — Clients photograph receipts in WeChat; bot ingests automatically
  2. ›Analyze — GPT-4o vision OCR + categorization (rent / payroll / meals)
  3. ›Notify — Anomalies routed to bookkeeper for review; clean records auto-filed
  4. ›Assign — QuickBooks line items pre-filled; bookkeeper approves in seconds

Year 1 ROI: 4.2× (consistent with industry benchmarks for professional services).

Next step: Graduated to L3-P3 deployment with QuickBooks + client portal integration.

Outcomes
Time per receipt
90 sec→5 sec
OCR accuracy
92%→99%
Owner overtime
4h / night→0
Tax-season clients
200→500+
Case 04·Beauty · Nail Salon

No-shows down 70%. Repeat customers nearly doubled. Front desk headcount cut in half.

Challenge

6 front-desk staff across 3 locations spent days on WeChat / IG booking coordination. No-show rate 25%. Repeat customer rate 30% (industry: 50%).

Workflow we shipped
  1. ›Receive — WeChat / IG / Xiaohongshu DMs unified into one booking queue
  2. ›Analyze — Customer profile (preferred artist / style / price tier)
  3. ›Notify — 24h + 2h reminder SMS in customer's language
  4. ›Assign — Birthday / holiday / dormant-90d auto-triggered re-engagement

+44% monthly revenue. Subscription cost: $499/month.

Next step: Moved to L3-P6 Salon Agent Subscription. 50% credit when they upgrade to custom deployment.

Outcomes
No-show rate
25%→8%
Front desk headcount
6→3
Repeat customer rate
30%→55%
Monthly revenue
$80K→$115K

Want your story here?

Ready to bring Forward Deployed Engineering to your business?

30-minute discovery call. We'll tell you in 5 minutes whether AI can save you money. If yes, we'll show you the path. If not, we'll save you the budget.