Awareness and Utilisation: The Plan’s First Eight Actions

What Is on the Agenda?

Two axes, eight actions. Four prepare, four deploy.

Be aware: people, data, rules

Four actions. The national literacy programme, the advanced talent programme, the data-driven push and the legal-ethical framework. The shared purpose is preparing the ground. Nothing here produces a product on its own; each makes the next thing possible. A team that does not know the tools cannot use them, a model cannot be built on disordered data, and nobody takes risk without rules.

Utilise: infrastructure, access, application

Four more. The national compute backbone, the GPU-for-All programme, public sector transformation and sectoral pilot programmes. These build on the prepared ground. Computing power, access channels and real application areas meet here. On the public side each ministry scans its own use cases; on the private side, sector pilots run.

What the eight have in common

All of them produce access. Actions in the other axes talk about investment, models and diplomacy. These eight open doors: a training door, a data door, a compute door, a pilot door. The doors are open, but going in requires an application. Nobody will knock and invite you. And applications require preparation — which is why the real work in front of each door is the work you do without waiting for it.

The calendar front-loads these two axes

The plan’s phasing puts these two first. Governance structures, data-space pilots, public applications and compute investment are prioritised in the early period. So these eight actions will move first. The plan also notes that early-maturing actions will scale without waiting for the calendar.

The other axes build on these two

Produce and Govern carry heavier instruments: a financing ladder, growth zones, model development calls. But all of them rest on these two axes. A business without capability and data cannot apply to a model fund. The order is not a preference but a requirement.

Why Is the Question Being Asked Now?

Four reasons make these two axes today’s subject.

The doors will open in sequence

Programmes do not start at once. Literacy comes early, the data library in the middle, large-scale credits later. A business that knows the order prepares in the same order. Trying to prepare for every door at once ends in being ready for none.

Preparation takes time

Team training takes weeks; data order takes months. Whoever begins these when a call opens is already late. Preparation is not an application requirement, but without it an application is useless. It also shows at the evaluation table: a prepared application and a rushed one do not look alike.

The economy favours small steps

Appetite for large investment is low. Most support in these two axes is small-scale: training, vouchers, credits, pilots. These are instruments suited to the current climate. Actions requiring large capital sit in later axes and later years.

Moving first pays

Quotas are limited and competition is thin in the first wave. First-wave references also help in later waves. A business that has completed one supported project applies more strongly for the second. A finished project is the most persuasive reference there is.

What Is Wrong?

Four assumptions cause businesses to miss these axes.

“Training and data can wait”

The most expensive error. If people and data are not ready, computing power is useless. That is exactly why the state put this axis first. Following the same order means following a tested path. A business that breaks the sequence cannot use the credit it receives; a strong engine goes nowhere with an empty tank.

“These supports are not built for our scale”

Most actions in these two axes look precisely at small scale. Free training, SME vouchers, separate access windows, ready-made solution packages. The plan targets at least a thousand SME vouchers. The distant axis is not this one; it is the eight actions talking about gigawatt investment and model export.

“The technical team will handle the data work”

Data classification is a commercial decision, not a technical one. Which data is core and which is shareable gets decided at the management table. It is a competitive advantage question, not a file format question. Without that distinction, teams are either too cautious or too generous.

“Let’s get the support first, then plan”

Reverse order. Building a project around a support produces work that ends when the voucher ends. The right order is the business need first, then the support that matches it. Projects built the other way are neither scaled nor closed; they are simply forgotten.

The Real Mechanism

The eight actions appear, on the business side, as four doors.

Door one: capability

The free literacy programme and the certified talent tracks. They let you build a three-level structure: safe user, applies it to their work, designs the process. Writing down who stands at which level stops you wasting training hours. Not everyone needs the third level.

Door two: data

The National Data Library and the Data Spaces programme. It works both ways: you can use public data and share your own. Both rest on the same precondition — your own data inventory in order.

Door three: compute

The GPU-for-All programme. GPU-hour credits open to startups and SMEs through separate windows, with differentiated packages for training and inference workloads. Projects get credit; curiosity does not. A one-sentence project definition is half the application.

Door four: application

Sectoral pilot programmes and SME vouchers. Multi-stage support running from idea to product: idea, prototype, field trial, scaling. Each stage eliminates and each stage increases support — project development at the idea stage, a pilot budget at the prototype stage, real-environment access in the field. The scan-pilot-scale loop is this door’s language.

Who Is Affected, and How?

Four profiles, four different priorities.

The micro business

Priority is door one. Free training makes a difference even in a one-person business. Second priority is the ready-made solution packages behind door four, designed for exactly this profile: easy to install, no technical knowledge required. The data and compute doors are further off — though drawing an inventory takes half a day even here, and will be needed later.

The growing SME

All four doors are open. But the order matters: capability and data first, compute and pilots second. This profile’s most common mistake is skipping the first two doors and applying straight to a pilot call. Even if accepted, the project stalls, because neither the team nor the data is ready.

The technology producer

Priority is doors three and four. GPU credits cut development costs and sector pilots supply the first reference. At least 40 percent of credits go to tech startups. The startup certificate is a threshold condition here; start that process early. The certificate does not wait for the call — the call waits for the certificate.

The consultant and training provider

These two axes create a market directly. Training for millions, voucher projects for thousands of businesses, data order for institutions. But competition will grow too. What will separate providers is knowing which door to take a client to. A client taken to the wrong door loses both money and trust.

Decision Order

Four steps, in door order.

One: draw the level ruler

Write a level and a date next to every name on the team. Half a day. Without this ruler no training plan works. Three levels are enough: safe user, applies it, designs it.

Two: draw the data inventory

Which data, where, in whose name, under which permission. Another half day. This table is the precondition for both using public data and sharing your own. It also surprises almost every business the first time: forgotten records, files held by one person.

Three: pick a pilot and write its criterion

Scan the processes, pick the one that gives the clearest result, write two numbers: where we are, where we want to be. The pilot starts without waiting for support. Two to four weeks is enough. Its aim is not to run but to teach.

Four: prepare the application file

Project sentence, data table, compute estimate, company profile. One day. Updating it when a call opens takes hours, and the same file works at different doors: vouchers, credits, pilot calls.

Where to Start?

Four weeks, four jobs.

Week one: the level ruler

Team list, three levels, target dates. Start with yourself; a subject the manager has not learned is not taken seriously by anyone else.

Week two: the data inventory

One table, sorted into four classes: core, shareable, personal, dead. Classification makes every later data decision easier.

Week three: the process scan

A one-hour meeting. Filter with three questions: does it repeat, does it eat time, is it measurable? Produce a five-line candidate list — with the people who do the work, because they know where the hours go.

Week four: first pilot and file

Pick a candidate and start the pilot. In the same week, build the skeleton of your application file. At month’s end you hold a working routine, not a plan.

What Not to Do?

Four traps.

Knocking on doors out of order

Compute credit is useless without data order. The sequence is capability, data, compute, application. Break it and you cannot use what you receive.

Sending everyone to the same training

Training assigned without level distinctions does not get completed. It also gets forgotten, because it touched nobody’s actual work.

Sharing data before classifying it

A share made without separating core from shareable can carry your competitive advantage outside. Classification comes before sharing.

Postponing the pilot until support arrives

The most common form of delay. A small pilot runs on your own budget and should run anyway. Support is for growing work whose experiment is finished.

What to Watch?

Four indicators suffice.

Training completion rate

How many people reached their target level? Completion counts, not enrolment. Training is easy to assign and hard to finish. The national programme uses the same measure: certificates depend on a minimum threshold.

Inventory freshness

When was the inventory last updated? Older than a year and it no longer reflects reality. New systems add new data types. An annual review is enough.

The pilot counter

How many pilots started this year, how many scaled, how many closed? All three are healthy outcomes. Zero pilots is the unhealthy one; nothing gets learned without trying.

Call tracking

Is someone checking quarterly? Which call opened, which did we apply to, which did we deliberately skip? The monitoring routine covers this habit.

How Does This Period End?

Three separations will become visible.

The gap between trained and untrained teams widens

Free training is open to everyone, which turns the gap into a decision difference rather than a budget one. The distance between those who use the opportunity and those who do not will be hard to close within a few years, because learning compounds: a team a year behind falls further behind in the second year.

Businesses with data order will pull ahead

As the data economy forms, the prepared will join it: using public data, valuing their own, drawing on incentives. Those waiting with scattered data will only watch. And ordering data is not an investment but a habit.

The doors will get crowded as they open

In the first wave quotas are generous and competition thin. As programmes become known, application numbers rise. The early-mover premium erodes over time. This curve repeats in every support programme.

A Solid Digital Foundation

Four stones, each one page.

The level ruler

Who stands where, by when. The basis of every training decision.

The classified data inventory

Four classes, one label per row. The foundation of both sharing and protection decisions.

The candidate process list

Five lines filtered by three questions. The source of pilot selection.

The application file skeleton

Project sentence, data table, compute estimate. Updated when a call opens, never written from scratch.

Frequently Asked Questions