Adoption

Everyone bought AI.
Few were changed by it.

The record on generic AI adoption is now public, and it is poor. Not because the intelligence is weak, but because of how it was adopted: generic, rented, cloud bound, and gone when the consultants left. Manar runs adoption the other way.

Start with a Site Read Open a conversation
The record

What the studies found

Three findings, from three independent sources, describe the same failure.

95%

of enterprise generative AI pilots returned nothing measurable. The authors' diagnosis was not model quality. Generic tools do not learn the operation they are dropped into.

MIT, The GenAI Divide: State of AI in Business, 2025

42%

of companies abandoned most of their AI initiatives before production, up from 17 percent a year earlier. The obstacles they named most: cost, data privacy, and security.

S&P Global Market Intelligence survey, 2025

280x

is how far the cost of generic AI intelligence fell in eighteen months. What everyone can rent for pennies is an advantage to no one.

Stanford HAI, AI Index Report, 2025

Read together: adoption is failing while the technology is succeeding. The failure is in the approach, and its causes are now known.

Why adoption fails

Four causes. Four reversals.

Across the published research, the same causes repeat. Manar's method was built against each of them.

01
It was generic.

A tool trained on the world in general knows nothing about your operation in particular, and it cannot learn it. Manar starts from a registered model and trains it on your data until it knows your conditions. Specific beats general. Always.

02
It was rented.

A subscription builds no asset. When it ends, nothing remains on your books but the invoices. A Manar deployment is owned from handover: the model, the hardware, and the growing record of your own operation.

03
It assumed the cloud.

The most cited obstacles in the abandonment data were cost, data privacy, and security. All three are properties of sending your data to someone else's computer. Manar's systems run inside your walls, offline if required, so the obstacle is removed rather than managed.

04
It left with the consultants.

Advice departs. Capability stays. A Manar engagement is complete only when your own staff operate the system without Manar in the room, and it keeps learning your operation after we leave.

The Adoption Path

Five stages. One direction.

Adoption at Manar is a defined path, not an open engagement. Each stage ends in something you can see, hold, or measure.

01 Site Read

A paid, scoped reading of your site: what runs, what sits unread, which registered model reads it first, and on what hardware. You receive a written finding either way.

02 First Deployment

Before work begins, we agree what will be measured and what number means acceptance. Then one registered model is trained on your data and installed on hardware you own. It is accepted when it meets the number, not when it is installed.

03 Handover

Your staff are trained on the running system. The stage is complete when they operate it without Manar in the room, demonstrated, not promised.

04 Registry Expansion

A new kind of work gets a model trained for it, on the same ground, reading what your systems already produce. Nothing is rebuilt from nothing.

05 Compound

The system retrains as your record grows. Accuracy on your conditions deepens with every month it runs, a record no later competitor can shortcut.

How Manar builds →
What arrives

Adoption that arrives in a box.

Manar deployments are hardware and software together. Stage one specifies the machine. Stage two arrives with it installed. Stage three hands you the keys. What remains on your site when the engagement closes:

Hardware you own A model that knows your ground Staff who run it A number it had to meet
The national direction

Inside the national direction

Pakistan's first National AI Policy, approved in July 2025, calls for local AI products, sectoral transformation, secure and ethical deployment, and one million trained professionals by 2030. Manar's work sits inside that direction: local products, deployed in Pakistani institutions, operated by Pakistani staff trained on the systems they own.

Manar and the National AI Policy →
Before you ask

Common questions

We already tried AI and it did not stick. Why would this be different? +

Then you are exactly who this path was designed for, and what you saw has now been measured across thousands of organizations. The pilots that failed were generic, rented, and cloud bound. This path reverses each of those properties, and stage two does not complete until the system meets a number we agree on your conditions before work begins.

Is this consulting? +

No. Every adoption engagement ends in a deployed, owned, measured system or it does not happen. Advice without deployment is not what Manar sells.

What does adoption cost? +

It begins with a Site Read, which is scoped and priced as a defined engagement. Deployment is priced on the finding, sized to what your operation actually needs.

Who owns everything at the end? +

You do. The model, the hardware it runs on, and every byte of the record it builds. Nothing is licensed back and nothing depends on Manar continuing to exist for your system to keep working.

Open a conversation.

Tell us what you run and what sits unread. We will tell you where adoption starts on your ground, and reply within two working days.

Send inquiry →
Continue