The blog
Field notes on AI agents in real organizations.
The actual stack, the real numbers, and what survives contact with production - from one Quick Win to a working agent layer, inside your environment.
The shift handover is your most expensive undocumented process
A plant running three shifts loses the same information every eight hours. Here is the use case, the business impact, and the four decisions you have to make before anyone writes code.
Read →
01
Onboarding thousands of frontline employees who do not have a computer
Nursing, security, facilities and logistics operators hire at volume, across the country, into roles with no desk. The training is a binder. Here is what an agent changes, and the four decisions that come first.
02
Field crews will send a voice note. They will not fill in your form
Refuelling, waste collection, infrastructure. The report from the field is a phone call or a scrap of paper, and the invoice argument happens weeks later. What an agent changes, and the four decisions behind it.
03
In distribution, the first quote back usually wins the order
Metals, chemicals, components. Every RFQ arrives in a different format and a person decodes it by hand. What an agent changes about response time, and the four decisions that keep it safe.
04
Batch records: collect the evidence during production, not the week the auditor arrives
Pharma and medical device manufacturers keep batch documentation in paper and spreadsheets. An agent can close the gap while it is still cheap to close. Here is where the line has to be drawn.
05
The Chrome extension that puts my CRM inside LinkedIn and WhatsApp
The full build, in the open: the one-function seam that makes a new platform a single file, the two LinkedIn IDs that are not interchangeable, the MV3 trap that quietly broke auto-reload, and why nothing it writes is anonymous.
06
MCP just became a real standard: what it means for connecting AI to your business systems
On 2026-07-28 the Model Context Protocol - the standard plug between AI agents and your existing systems - shipped its biggest revision yet. Here is why that quiet update lowers the cost and risk of a first automation.
07
How long until an AI agent pays for itself? The 2026 payback numbers
The 2026 data is in: the median AI agent pays for itself in about 5 months, but it splits sharply by function. Here is what that means for a smaller business - and how to land on the fast side.
08
88% of companies running AI agents had a security incident last year - what it means for your business
Two 2026 surveys found about 88% of organizations running AI agents hit a security incident in the past year. Here is exactly what a smaller business should demand before turning one on.
09
The AI project graveyard: why 42% of companies walked away in 2025 - and how your business avoids it
In 2025, 42% of companies abandoned most of their AI initiatives, up from 17%. The projects that die are scoped and governed badly - here is how a smaller business ships one that lasts.
10
Claude Opus 5 took the #1 spot: why the newest AI model is not your bottleneck
A new top-ranked model landed on July 24, 2026. For a traditional or small business the frontier model was never the constraint - here is what actually decides AI ROI.
11
AI just got cheaper: what the 2026 price cuts mean for automating your business
The 2026 models got cheaper and more reliable at once. Here is how falling model prices change which of your everyday processes are now worth automating.
12
AI agent sprawl: the enterprise mess your business can skip
96% of enterprises run AI agents and 94% fear the sprawl. Here is what it means for a smaller company - and how to design around it from day one.
13
AI agent trends in mid-2026: what actually matters for small and traditional businesses
Workflow agents over chatbots, multi-agent teams, and the ROI gap - what each trend means in practice, with the numbers behind it.
14
AI agents for small businesses: the practical playbook (2026)
Where to start this week, what to automate first, DIY vs a partner, what it costs - and the classic mistake to avoid.
15
Implementing Claude in your organization: a practical guide for Israeli companies (2026)
Deployment options (API, Bedrock, Vertex), how to choose an implementation partner, what small businesses should do differently, and what it costs.
16
Back-office automation with AI agents: the complete guide (2026)
What agents do that RPA can't, which processes to automate first, what the data says, and how to deploy inside your environment.
17
AI agents for traditional industries: why you stand to gain the most
Factories, logistics and professional firms - that's where agents return the most. On your existing systems, replacing nothing, data never leaves.
18
My real AI sales stack as a solo founder (2026)
The exact tools behind a one-person GTM team: Claude Code, a self-hosted CRM, Fathom, Apollo, WhatsApp automation - and the principles that make it safe.
19
Why most AI agents never reach production
Over 80% of agents stall between the demo and the floor. The five reasons we see again and again - and what to do differently.
20
LinkedIn outreach in 2026: what actually worked for me
~212 targeted connection requests, zero warnings, real meetings. The volume caps, skip logic and warm-content playbook.
21
The 60-second voice habit that keeps my CRM actually updated
CRMs die because updating them is friction. I record one minute of speech after every meeting - an agent does the typing.
22
Deploying AI without your data ever leaving: a guide for security owners
The question that stops projects is 'where does the data go?'. An in-your-environment architecture turns security from blocker to engine.
23
Implementing AI agents in your organization - the full guide
Five steps from the field: find the bottleneck, scope a Quick Win, run inside your environment, human oversight, measure ROI.
24
Custom AI agent or off-the-shelf copilot? How to decide
When is Copilot enough, and when do you need an agent built on your process? A simple decision framework.
25
I never walk into a sales call unprepared - an agent does the prep
Twice a day an agent scans my calendar and files a one-page research brief for every external meeting.
02
Field crews will send a voice note. They will not fill in your form
Refuelling, waste collection, infrastructure. The report from the field is a phone call or a scrap of paper, and the invoice argument happens weeks later. What an agent changes, and the four decisions behind it.
03
In distribution, the first quote back usually wins the order
Metals, chemicals, components. Every RFQ arrives in a different format and a person decodes it by hand. What an agent changes about response time, and the four decisions that keep it safe.
04
Batch records: collect the evidence during production, not the week the auditor arrives
Pharma and medical device manufacturers keep batch documentation in paper and spreadsheets. An agent can close the gap while it is still cheap to close. Here is where the line has to be drawn.
05
The Chrome extension that puts my CRM inside LinkedIn and WhatsApp
The full build, in the open: the one-function seam that makes a new platform a single file, the two LinkedIn IDs that are not interchangeable, the MV3 trap that quietly broke auto-reload, and why nothing it writes is anonymous.
06
MCP just became a real standard: what it means for connecting AI to your business systems
On 2026-07-28 the Model Context Protocol - the standard plug between AI agents and your existing systems - shipped its biggest revision yet. Here is why that quiet update lowers the cost and risk of a first automation.
07
How long until an AI agent pays for itself? The 2026 payback numbers
The 2026 data is in: the median AI agent pays for itself in about 5 months, but it splits sharply by function. Here is what that means for a smaller business - and how to land on the fast side.
08
88% of companies running AI agents had a security incident last year - what it means for your business
Two 2026 surveys found about 88% of organizations running AI agents hit a security incident in the past year. Here is exactly what a smaller business should demand before turning one on.
09
The AI project graveyard: why 42% of companies walked away in 2025 - and how your business avoids it
In 2025, 42% of companies abandoned most of their AI initiatives, up from 17%. The projects that die are scoped and governed badly - here is how a smaller business ships one that lasts.
10
Claude Opus 5 took the #1 spot: why the newest AI model is not your bottleneck
A new top-ranked model landed on July 24, 2026. For a traditional or small business the frontier model was never the constraint - here is what actually decides AI ROI.
11
AI just got cheaper: what the 2026 price cuts mean for automating your business
The 2026 models got cheaper and more reliable at once. Here is how falling model prices change which of your everyday processes are now worth automating.
12
AI agent sprawl: the enterprise mess your business can skip
96% of enterprises run AI agents and 94% fear the sprawl. Here is what it means for a smaller company - and how to design around it from day one.
13
AI agent trends in mid-2026: what actually matters for small and traditional businesses
Workflow agents over chatbots, multi-agent teams, and the ROI gap - what each trend means in practice, with the numbers behind it.
14
AI agents for small businesses: the practical playbook (2026)
Where to start this week, what to automate first, DIY vs a partner, what it costs - and the classic mistake to avoid.
15
Implementing Claude in your organization: a practical guide for Israeli companies (2026)
Deployment options (API, Bedrock, Vertex), how to choose an implementation partner, what small businesses should do differently, and what it costs.
16
Back-office automation with AI agents: the complete guide (2026)
What agents do that RPA can't, which processes to automate first, what the data says, and how to deploy inside your environment.
17
AI agents for traditional industries: why you stand to gain the most
Factories, logistics and professional firms - that's where agents return the most. On your existing systems, replacing nothing, data never leaves.
18
My real AI sales stack as a solo founder (2026)
The exact tools behind a one-person GTM team: Claude Code, a self-hosted CRM, Fathom, Apollo, WhatsApp automation - and the principles that make it safe.
19
Why most AI agents never reach production
Over 80% of agents stall between the demo and the floor. The five reasons we see again and again - and what to do differently.
20
LinkedIn outreach in 2026: what actually worked for me
~212 targeted connection requests, zero warnings, real meetings. The volume caps, skip logic and warm-content playbook.
21
The 60-second voice habit that keeps my CRM actually updated
CRMs die because updating them is friction. I record one minute of speech after every meeting - an agent does the typing.
22
Deploying AI without your data ever leaving: a guide for security owners
The question that stops projects is 'where does the data go?'. An in-your-environment architecture turns security from blocker to engine.
23
Implementing AI agents in your organization - the full guide
Five steps from the field: find the bottleneck, scope a Quick Win, run inside your environment, human oversight, measure ROI.
24
Custom AI agent or off-the-shelf copilot? How to decide
When is Copilot enough, and when do you need an agent built on your process? A simple decision framework.
25
I never walk into a sales call unprepared - an agent does the prep
Twice a day an agent scans my calendar and files a one-page research brief for every external meeting.