MyPostMYPOST
Blog

Notes from inside the deployments

What works, what breaks, and what we'd do differently — written for the people who have to run the systems.

Amazon Vendor Central: EDI Order Processing From Seven Days to Ten Minutes
Case Study
June 18, 20269 min read

Amazon Vendor Central: EDI Order Processing From Seven Days to Ten Minutes

A supplier was losing a week to every Amazon EDI batch, and losing chargebacks to the delay. The fix was less about AI cleverness than about never letting a malformed document reach a human.

Read the case study
Replacing Legacy Enterprise Search With a Retrieval-Grounded AI Agent
Case Study
May 29, 20268 min read

Replacing Legacy Enterprise Search With a Retrieval-Grounded AI Agent

Keyword search was returning documents; people needed answers. The engineering was less about the model than about making sure every sentence it produced could be traced to a source.

Read the case study
Stable Diffusion XL: 73% Faster Inference Without Losing Image Quality
Case Study
May 7, 20267 min read

Stable Diffusion XL: 73% Faster Inference Without Losing Image Quality

2.58 seconds to 0.7 seconds per 1024×1024 image on consumer GPUs. The hard part was proving the faster pipeline was still producing the same pictures.

Read the case study
An Internal AI Builder That Cut Deployment Cycles 60%
Case Study
April 22, 20268 min read

An Internal AI Builder That Cut Deployment Cycles 60%

Every new automation was a bespoke engineering project, so most never got built. The fix was a platform that let non-technical teams ship their own — with governance that made that safe.

Read the case study
Support AI Agents: 57% Fewer Tickets and an 18-Point CSAT Gain
Case Study
April 3, 20268 min read

Support AI Agents: 57% Fewer Tickets and an 18-Point CSAT Gain

Deflection is easy to fake and expensive to get wrong. Routing to specialised agents — and knowing when to hand to a human — did more for satisfaction than any single model choice.

Read the case study
Lead Scoring and Outreach at 12,000 Leads a Week, With Reply Rates Up 42%
Case Study
March 19, 20268 min read

Lead Scoring and Outreach at 12,000 Leads a Week, With Reply Rates Up 42%

Enrichment and scoring are the easy half. The reply rate moved because the messages stopped sounding like they were generated, which took more work than the model did.

Read the case study
E-commerce Multi-Agent System: AOV Up 22%, Abandoned Carts Down 34%
Case Study
February 26, 20268 min read

E-commerce Multi-Agent System: AOV Up 22%, Abandoned Carts Down 34%

Cart recovery, personalisation and upsells are usually three disconnected tools sending three uncoordinated messages. Running them as agents over one behavioural model is what changed the numbers.

Read the case study
250,000 Documents, 94% Retrieval Accuracy, 75% Less Research Time
Case Study
February 5, 20269 min read

250,000 Documents, 94% Retrieval Accuracy, 75% Less Research Time

At a quarter of a million documents, retrieval strategy stops being a detail and becomes the entire product. Here is what we measured, and what we refused to promise.

Read the case study
Revenue Automation Agents: 8,000 Leads a Month, Conversions Up 31%
Case Study
January 21, 20268 min read

Revenue Automation Agents: 8,000 Leads a Month, Conversions Up 31%

Qualification, proposals, scoring and follow-up ran as four disconnected manual processes. Wiring them into one agent chain removed 68% of the manual load without removing the judgement.

Read the case study
A Research Agent That Reads 100+ Long-Form Sources Per Run
Case Study
January 8, 20268 min read

A Research Agent That Reads 100+ Long-Form Sources Per Run

Not a GPT wrapper — a research analyst. Citation-backed briefing documents from podcasts, transcripts, blogs and video, with factual consistency measured rather than assumed.

Read the case study
Real-Time Voice Coaching: Holding a Conversation Inside 2.5 Seconds
Case Study
December 11, 20257 min read

Real-Time Voice Coaching: Holding a Conversation Inside 2.5 Seconds

Speech to text to model to speech, in a loop tight enough that users forget they are talking to software. Every architectural decision was made against a latency budget.

Read the case study
SOC 2 and GDPR Intake: From Days to Under Two Minutes
Case Study
November 20, 20257 min read

SOC 2 and GDPR Intake: From Days to Under Two Minutes

Compliance enquiries were being lost to response time, not to price. Automating 70–90% of intake meant the qualified ones reached a human while they still cared.

Read the case study
Outbound at 5–10× Throughput With Reply Rates Up, Not Down
Case Study
October 30, 20258 min read

Outbound at 5–10× Throughput With Reply Rates Up, Not Down

Most outbound automation trades reply rate for volume. This one raised both, because the research step was automated properly rather than skipped.

Read the case study
Facial Recognition: Precision Up 10–25%, False Positives Down 30–60%
Case Study
October 9, 20257 min read

Facial Recognition: Precision Up 10–25%, False Positives Down 30–60%

In identity verification, a false positive is not a rounding error. Most of the gain came from preprocessing and threshold work, not from a bigger model.

Read the case study
Social Sentiment at Scale: A Pipeline That Survives Real Text
Case Study
September 18, 20256 min read

Social Sentiment at Scale: A Pipeline That Survives Real Text

Sentiment models look easy until they meet sarcasm, spam and emoji. Most of the accuracy came from what we removed before classification, not from the classifier.

Read the case study
Predicting Restaurant Tips: A Regression Pipeline That Beat the Baseline 15–30%
Case Study
August 27, 20256 min read

Predicting Restaurant Tips: A Regression Pipeline That Beat the Baseline 15–30%

A small, honest modelling problem done properly: feature engineering, cross-validation, and a baseline that had to be beaten before anything shipped.

Read the case study
Forecasting Sales From Marketing Spend, 20–35% More Reliably
Case Study
August 6, 20257 min read

Forecasting Sales From Marketing Spend, 20–35% More Reliably

Budget was being allocated on last quarter's results and instinct. Modelling the channels together produced forecasts that were both more reliable and interpretable enough to act on.

Read the case study
Payment Fraud Screening: 90%+ Accuracy With 40–70% Less Manual Review
Case Study
July 16, 20257 min read

Payment Fraud Screening: 90%+ Accuracy With 40–70% Less Manual Review

Fraud detection is a threshold problem wearing a modelling problem's clothes. Getting the operating point right mattered more than the choice of classifier.

Read the case study
LSTM Time-Series Forecasting, and an Honest Account of What It Can Do
Case Study
June 25, 20256 min read

LSTM Time-Series Forecasting, and an Honest Account of What It Can Do

Sequence modelling on market data, evaluated against a naive baseline rather than against a nice-looking chart. The caveats are the most useful part of this one.

Read the case study
The Future of AI Automation: How It's Changing Business Operations
Article
March 12, 20256 min read

The Future of AI Automation: How It's Changing Business Operations

Automation stopped being a cost-cutting exercise and became the operating layer businesses are built on. Here's what that shift looks like from inside the companies making it.

Read article
5 Must-Have AI Tools to Streamline Your Business Tasks
Resources
March 4, 20255 min read

5 Must-Have AI Tools to Streamline Your Business Tasks

A short, opinionated list of the categories worth your budget this year — and the questions to ask before you sign anything.

Read article
AI vs. Manual Work: Which One Saves More Time & Money?
Article
February 21, 20257 min read

AI vs. Manual Work: Which One Saves More Time & Money?

The honest answer is that it depends on volume, variance and error cost. Here's a framework for working out which side of the line your process falls on.

Read article
How AI is Transforming Workflow Automation for Businesses
Article
February 8, 20256 min read

How AI is Transforming Workflow Automation for Businesses

Rule-based automation broke the moment reality deviated from the flowchart. Models that handle ambiguity changed which processes are worth automating at all.

Read article

Find out what's automatable in 30 minutes

Bring one process that's eating your week. You leave with a prioritized list and a rough cost — whether or not you hire us.

Book a free audit