Greg Garnet

AI Automation and Optimization in Business

[Greg Garnet @GarnetGreg]

An insurer cut underwriting cost up to 80% and moved decisions from days to hours.

The AI was built in 60 days.

EXL's assistant reads hundreds of pages, extracts evidence, and prepares an initial assessment for a human underwriter.

That changes the economics twice:

1. lower cost per policy reviewed
2. more policies reviewed with the same team

AWS says early adoption also increased signed customer deals for EXL.

The 80% figure is based on EXL analysis and described as “up to,” not an audited average.

The winning workflow keeps judgment human.

It removes the document hunt before judgment begins.

[Greg Garnet @GarnetGreg]

One sales team moved lead conversion from 19% to 78%.

That is more than 4x — without claiming the AI wrote magical emails.

ARNOLD fixed the plumbing:

- centralized webinar and trade-show leads
- identified high-value contacts earlier
- automated assignment
- exposed context inside the seller's workflow
- tightened marketing-to-sales follow-up

Microsoft reports the conversion change and a shorter sales cycle.

Copilot was one component of a broader Dynamics and Power Platform redesign.

Most companies blame weak leads.

Sometimes the real problem is that a strong lead reaches the right salesperson three days too late.

[Greg Garnet @GarnetGreg]

A grocery retailer lost roughly 20% of overall sales — while AI-managed trial products held their ground.

That gap is the story.

Super Hosokawa used demand forecasting to decide:

- what to order
- how much to stock
- which items risked waste
- where replenishment needed intervention

Microsoft describes the relative performance as an effective 20% sales improvement for trial items.

That is not the same as saying total company revenue rose 20%.

The lesson for every inventory business:

AI does not need to predict the future perfectly.

It only needs to make fewer expensive stock decisions than the current process.

[Greg Garnet @GarnetGreg]

One sales team moved lead conversion from 19% to 78%.

That is more than 4x — without claiming the AI wrote magical emails.

ARNOLD fixed the plumbing:

- centralized webinar and trade-show leads
- identified high-value contacts earlier
- automated assignment
- exposed context inside the seller's workflow
- tightened marketing-to-sales follow-up

Microsoft reports the conversion change and a shorter sales cycle.

Copilot was one component of a broader Dynamics and Power Platform redesign.

Most companies blame weak leads.

Sometimes the real problem is that a strong lead reaches the right salesperson three days too late.

[Greg Garnet @GarnetGreg]

An insurer cut underwriting cost up to 80% and moved decisions from days to hours.

The AI was built in 60 days.

EXL's assistant reads hundreds of pages, extracts evidence, and prepares an initial assessment for a human underwriter.

That changes the economics twice:

1. lower cost per policy reviewed
2. more policies reviewed with the same team

AWS says early adoption also increased signed customer deals for EXL.

The 80% figure is based on EXL analysis and described as “up to,” not an audited average.

The winning workflow keeps judgment human.

It removes the document hunt before judgment begins.

[Greg Garnet @GarnetGreg]

A retailer turned 32 minutes of product hunting into 1–2 minutes — then lifted conversion 10%.

Central Group gave personal shoppers visual AI search.

A customer sends a picture.

The system finds matching products, current prices, promotions, and stock.

The shopper gets an answer before purchase intent dies.

Google Cloud reports a 94% reduction in search time alongside the conversion lift.

The public case does not isolate every factor behind conversion.

Still, the mechanism is clear:

customers do not abandon because they hate your product.

They abandon because finding it became work.

[Greg Garnet @GarnetGreg]

An industrial marketplace generated 500+ orders worth ₹2.1M from one AI commerce assistant.

Then a second AI workflow grew a sourcing operation from roughly ₹12 crore to ₹50 crore per quarter.

Moglix connected 550 new suppliers and automated product discovery across 700,000+ industrial items.

The stack attacked 3 leaks:

- buyers could not find exact specifications
- teams could not find the right vendor
- PDF product data stayed trapped in documents

Google Cloud reports the figures; concurrent growth factors are not isolated.

The opportunity is not “add a chatbot.”

It is turning an unanswered product request into an order before the buyer goes elsewhere.

Comment LEAKS. (also RT + Like)

and I will DM you the complete AI Revenue Recovery Kit: the 15-leak audit, editable ROI scorecard, guardrails, and 7-day pilot plan.

[Greg Garnet @GarnetGreg]

A tiny climate-tech company doubled revenue while making AI iteration 10x faster.

Perennial also processed 10x more data and expanded its monitored footprint 5x.

Its product measures soil carbon remotely for agriculture markets.

That means growth depended on one constraint:

how much land the system could analyze accurately.

Google Cloud reports the 2023-to-2024 revenue doubling.

Perennial projected another 5x in 2025, but a projection is not a realized result.

The reusable play:

find the production bottleneck that caps how many customers you can serve.

Then apply AI there — not to another internal summary.

[Greg Garnet @GarnetGreg]

An industrial marketplace generated 500+ orders worth ₹2.1M from one AI commerce assistant.

Then a second AI workflow grew a sourcing operation from roughly ₹12 crore to ₹50 crore per quarter.

Moglix connected 550 new suppliers and automated product discovery across 700,000+ industrial items.

The stack attacked 3 leaks:

- buyers could not find exact specifications
- teams could not find the right vendor
- PDF product data stayed trapped in documents

Google Cloud reports the figures; concurrent growth factors are not isolated.

The opportunity is not “add a chatbot.”

It is turning an unanswered product request into an order before the buyer goes elsewhere.

Comment LEAKS. (also RT + Like)

and I will DM you the complete AI Revenue Recovery Kit: the 15-leak audit, editable ROI scorecard, guardrails, and 7-day pilot plan.

[Greg Garnet @GarnetGreg]

A tiny climate-tech company doubled revenue while making AI iteration 10x faster.

Perennial also processed 10x more data and expanded its monitored footprint 5x.

Its product measures soil carbon remotely for agriculture markets.

That means growth depended on one constraint:

how much land the system could analyze accurately.

Google Cloud reports the 2023-to-2024 revenue doubling.

Perennial projected another 5x in 2025, but a projection is not a realized result.

The reusable play:

find the production bottleneck that caps how many customers you can serve.

Then apply AI there — not to another internal summary.

[Greg Garnet @GarnetGreg]

One construction startup cut proposal time 80% and increased gross margin 20%.

The proposal did not become shorter.

The waiting did.

ICG used AI across the revenue workflow:

1. retrieve project knowledge
2. draft the response
3. price around value
4. simplify billing
5. reuse the system for client work

Microsoft also reports $500K in custom programming costs avoided.

These are company-reported results, and several workflow changes happened together.

But every slow proposal creates the same invisible tax:

the buyer's urgency decays while your team formats documents.

[Greg Garnet @GarnetGreg]

$7.2M in new airline revenue came from a place most companies still treat like support: WhatsApp.

IndiGo's AI channel served 3M+ users and issued 1M+ boarding passes.

But the clever part was the revenue path:

- a meal question becomes an add-on
- a seat question becomes an upgrade
- a flight alert becomes a hotel or insurance offer
- the purchase stays inside the conversation

Google Cloud reports ₹60+ crore from bookings, upgrades, and ancillaries.

That is a vendor customer story, not an independent audit.

The lesson is still brutal:

if your bot can answer but cannot transact, you built a cost center where a sales channel could exist.

[Greg Garnet @GarnetGreg]

A retailer turned 32 minutes of product hunting into 1–2 minutes — then lifted conversion 10%.

Central Group gave personal shoppers visual AI search.

A customer sends a picture.

The system finds matching products, current prices, promotions, and stock.

The shopper gets an answer before purchase intent dies.

Google Cloud reports a 94% reduction in search time alongside the conversion lift.

The public case does not isolate every factor behind conversion.

Still, the mechanism is clear:

customers do not abandon because they hate your product.

They abandon because finding it became work.

[Greg Garnet @GarnetGreg]

One construction startup cut proposal time 80% and increased gross margin 20%.

The proposal did not become shorter.

The waiting did.

ICG used AI across the revenue workflow:

1. retrieve project knowledge
2. draft the response
3. price around value
4. simplify billing
5. reuse the system for client work

Microsoft also reports $500K in custom programming costs avoided.

These are company-reported results, and several workflow changes happened together.

But every slow proposal creates the same invisible tax:

the buyer's urgency decays while your team formats documents.

[Greg Garnet @GarnetGreg]

$7.2M in new airline revenue came from a place most companies still treat like support: WhatsApp.

IndiGo's AI channel served 3M+ users and issued 1M+ boarding passes.

But the clever part was the revenue path:

- a meal question becomes an add-on
- a seat question becomes an upgrade
- a flight alert becomes a hotel or insurance offer
- the purchase stays inside the conversation

Google Cloud reports ₹60+ crore from bookings, upgrades, and ancillaries.

That is a vendor customer story, not an independent audit.

The lesson is still brutal:

if your bot can answer but cannot transact, you built a cost center where a sales channel could exist.

[Greg Garnet @GarnetGreg]

3 people listed 300 SKUs a day. AI cut the team to 2 and the cycle to 1-2 days.

Translation cost also fell 40%.

TVCMALL had 1M+ products and constant multilingual updates.

The new pipeline:

1. extracts messy supplier data
2. standardizes the product fields
3. translates the page
4. checks structured output
5. publishes across markets

AWS says the solution launched in one month and listing efficiency improved 30%.

Those are company-reported figures.

The money is not in generating more words.

It is in removing the queue between inventory arriving and inventory becoming sellable.

[Greg Garnet @GarnetGreg]

One AI-assisted format can produce up to 10x the revenue of the original article.

Meanwhile, 53% of small US news outlets fear they may not survive 5 years.

Nota turns a written story into distribution assets, including video.

The mechanism is painfully simple:

- research once
- write once
- adapt for each platform
- sell higher-value inventory
- keep journalists on reporting

Microsoft published the case; “up to 10x” is a ceiling, not a guaranteed average.

For agencies and expert businesses, the question is not “Can AI write?”

It is: how many sellable formats are trapped inside work you already paid to create?

Comment SCORECARD. (also RT + Like)

and I will DM you the complete AI Revenue Recovery Kit: the content-value audit, editable ROI scorecard, 15 revenue-leak models, and 7-day pilot plan.

[Greg Garnet @GarnetGreg]

An AI demo lab was followed by 139% sales growth and 131% more partner transactions.

It expanded into 23 countries.

TD SYNNEX stopped explaining the product with slides.

It let prospects use their own data inside a temporary lab:

- ask natural-language questions
- build dashboards
- automate workflows
- test a real use case
- delete the environment afterward

AWS reports the year-over-year Latin America figures; other factors were not isolated.

The B2B lesson is bigger than the tool:

when a prospect experiences value before procurement, the demo becomes part of the sales engine.

[Greg Garnet @GarnetGreg]

One AI-assisted format can produce up to 10x the revenue of the original article.

Meanwhile, 53% of small US news outlets fear they may not survive 5 years.

Nota turns a written story into distribution assets, including video.

The mechanism is painfully simple:

- research once
- write once
- adapt for each platform
- sell higher-value inventory
- keep journalists on reporting

Microsoft published the case; “up to 10x” is a ceiling, not a guaranteed average.

For agencies and expert businesses, the question is not “Can AI write?”

It is: how many sellable formats are trapped inside work you already paid to create?

Comment SCORECARD. (also RT + Like)

and I will DM you the complete AI Revenue Recovery Kit: the content-value audit, editable ROI scorecard, 15 revenue-leak models, and 7-day pilot plan.

[Greg Garnet @GarnetGreg]

3 people listed 300 SKUs a day. AI cut the team to 2 and the cycle to 1-2 days.

Translation cost also fell 40%.

TVCMALL had 1M+ products and constant multilingual updates.

The new pipeline:

1. extracts messy supplier data
2. standardizes the product fields
3. translates the page
4. checks structured output
5. publishes across markets

AWS says the solution launched in one month and listing efficiency improved 30%.

Those are company-reported figures.

The money is not in generating more words.

It is in removing the queue between inventory arriving and inventory becoming sellable.