When a Company Says It Allows AI

Published On: June 28th, 2026Last Updated: September 14th, 20264939 words24.7 min readDaily Views: 2Total Views: 17

A company can announce that it allows employees to use generative AI in a single sentence.

That does not mean the company is ready for what follows.

Permission answers only one question:

Are staff allowed to use these tools?

It does not answer:

  • Which tools are appropriate for which work?
  • What information may be entered into them?
  • What does an acceptable visual request look like?
  • Who is responsible for the decisions inside the result?
  • How should employees apply the company’s visual identity?
  • What happens when staff have very different levels of experience?
  • When does an AI-generated draft still require a designer?
  • Who reviews the final asset before it represents the organisation?

A company may be open to AI while having no shared understanding of how AI-assisted work should actually be conducted.

That gap creates problems which are often described as prompting problems, design problems, staff-performance problems, or even problems with the models themselves.

Sometimes they are all of those things.

Often, however, the deeper problem is that the organisation granted access to a new production capability without creating a shared working method around it.

The Visual Register by Mithaq Praxis is not intended to solve the whole of corporate AI governance.

It is being built to address one narrower part of the problem:

How does an organisation establish a shared visual source of truth and turn individual image requests into briefs that people, designers, and generators can understand?

Permission is not preparation

When a company says that AI is allowed, employees may hear very different things.

One person hears:

I can use this to help draft ideas.

Another hears:

I can now produce all the graphics I need without involving the design team.

A manager may hear:

Visual work should take less time and cost less.

A designer may hear:

I will now receive more low-quality material that I am expected to repair quickly.

An IT or security team may hear:

Staff are about to enter company information into tools we have not properly evaluated.

A brand lead may hear:

Every department can now generate its own interpretation of the company.

All of these reactions can emerge from the same announcement.

The phrase AI is allowed sounds like a policy.
In practice, it may be no more than a permission slip.
It says nothing about competence, process, responsibility, or quality.

The organisation has opened a door without deciding what people are expected to do after walking through it.

The same tool does not create the same capability

Two employees can use the same image generator and perform entirely different kinds of work.

One may write:

Make a futuristic office image for our website.

Another may define:

  • the communication objective;
  • intended audience;
  • type of workplace;
  • number and roles of the people shown;
  • specific activity;
  • approved product or technology;
  • visual hierarchy;
  • brand palette;
  • lighting;
  • composition;
  • text placement;
  • restricted details;
  • reasons the scene belongs to the company.

Both employees may technically be “using AI.”
They are not carrying out equivalent processes.

  • The first request leaves nearly every visual decision to the model.
  • The second attempts to direct those decisions before generation.

The difference is not merely prompt length.

  • A long prompt can still be incoherent.
  • A short prompt can sometimes be precise.

The difference lies in whether the employee understands that an image contains decisions and accepts responsibility for making the ones relevant to the work.

Generative tools can hide this difference because both users may receive polished-looking outputs.

A weakly directed image can still contain:

  • realistic lighting;
  • impressive architecture;
  • attractive people;
  • sophisticated colour grading;
  • plausible surface detail.

The polish may disguise the fact that the scene does not communicate what the company needs.
A model can generate visual confidence without organisational accuracy.

That is one reason companies may underestimate the gap between employees who understand visual direction and employees who only know how to request an image.

Prompting skill is uneven, but that is not the whole problem

It is tempting to describe this as a training issue:

Some staff are good at prompting and others are not.

There is truth in that.

People differ in:

  • technical vocabulary;
  • visual literacy;
  • writing ability;
  • understanding of image models;
  • awareness of composition;
  • familiarity with brand systems;
  • ability to analyse why an output failed.

But reducing the problem to individual prompting ability allows the company to avoid examining its own missing infrastructure.
Even a highly skilled employee cannot reliably apply a visual identity that has never been defined for generative use.

They may have access to a conventional brand guide containing:

  • logos;
  • colour codes;
  • fonts;
  • layout examples;
  • approved photography.

That guide may not answer:

  • What should our generated office environments look like?
  • Which aspects of our real workplace must remain accurate?
  • How should technology be depicted?
  • Should generated people appear in public-facing materials?
  • Which cultural or demographic stereotypes must be actively avoided?
  • What visual metaphors have become clichés in our industry?
  • How much fictionalisation is acceptable?
  • Which products, uniforms, tools, or safety equipment must be exact?
  • What information must never be submitted to a public model?
  • Which image types require human design review?

An employee cannot retrieve answers the company has not stored.
Without a shared visual system, every user becomes an accidental brand strategist.

  • Some will make thoughtful decisions.
  • Some will imitate whatever the model associates with words such as innovative, professional, premium, or futuristic.

The resulting inconsistency is not only a personal failure.

It is an organisational design failure.

Generators fill every silence

When a brief is vague, the generator does not leave the missing decisions blank.

It fills them.

Ask for a futuristic corporate office and a model may supply:

  • blue holograms;
  • glass walls;
  • dramatic city views;
  • anonymous employees;
  • spotless white furniture;
  • glowing interfaces;
  • no visible paperwork;
  • no evidence of the company’s actual work.

Ask for a diverse team and the model may create a carefully distributed group of stock-photo identities standing around a laptop, all smiling at the same angle.
Ask for an innovative laboratory and it may produce decorative science fiction rather than credible equipment.
Ask for a premium product image and it may place the item on reflective black stone beneath theatrical lighting, regardless of whether that visual language suits the company.

These are not random mistakes.
They are common visual associations produced where the organisation supplied no stronger direction.

The model fills the silence with what it has seen most often.

A company that does not define itself clearly will be represented through the generator’s defaults.
This can create a strange situation in which different departments produce polished visuals that resemble different fictional companies.

  • One team appears to work inside a neon technology corporation.
  • Another appears to belong to a soft beige wellness brand.
  • Another uses generic flat-vector illustrations.
  • Another generates glossy stock photography.

Each output may be individually competent.
Together, they do not form an organisation.

A brand guide is not always an operational visual system

Traditional brand guides remain important.
The Visual Register is not intended to replace them.

But many were designed for human designers who already understand how to interpret visual systems.

A brand guide may say:

Use our primary green as an accent.

  • A designer understands that this does not require every wall, shirt, chair, screen, and folder to be green.
  • A generator may need more explicit direction.
  • A brand guide may show several photographs of employees in natural office environments.

A designer can infer:

  • candid rather than posed interaction;
  • moderate environmental detail;
  • realistic daylight;
  • warm but professional expressions;
  • no exaggerated corporate theatre.

A non-designer may see only:

We use photos of people at work.

They may request a scene in which six people point enthusiastically at glowing graphs while standing inside an implausibly pristine boardroom.

The problem is not that the brand guide failed.
It was created for a context where human visual judgment was assumed.

Generative workflows expose how much interpretation designers perform between a written standard and a finished image.
A Company Visual Bible can extend the brand guide into a more operational source of truth.

It can state not only what assets exist, but how the organisation should be visually interpreted when a new scene is generated.

The company benefits from creating the Bible

For independent creators, a Visual Bible may preserve a fictional world, recurring characters, environments, attire, motifs, and artistic identity.

For a company, the Bible serves another purpose:

It prevents the organisation’s visual identity from being reconstructed independently inside every employee’s prompt.

A Company Visual Bible might contain:

  • brand purpose and visual posture;
  • approved logos and wordmarks;
  • palette and material language;
  • workplace environments;
  • product and equipment references;
  • staff attire or safety requirements;
  • preferred photographic and illustrative treatments;
  • principles for representing customers and employees;
  • standard image formats;
  • text-space requirements;
  • accessibility considerations;
  • restricted or confidential elements;
  • common unwanted clichés;
  • review requirements;
  • explanations of why important rules exist.

The organisation makes these decisions once and revises them deliberately.

Employees inherit the relevant parts when creating a brief.
They are not expected to remember the entire visual system or paste a brand manual into a prompt.

This does not remove their responsibility.
It relocates it.

  • The brand team decides what should remain.
  • The requester decides what this image needs to communicate.
  • The designer or creator decides how that communication becomes a workable visual concept.
  • The reviewer decides whether the result is suitable for release.
  • The generator interprets the directions it receives.

That is a more honest distribution of responsibility than asking everyone to “use AI” and treating every output as equivalent.

The employee must still supply the present need

A strong Company Visual Bible cannot compensate for an absent brief.

A manager may request:

Make something for recruitment.

The Bible can provide:

  • brand palette;
  • workplace appearance;
  • approved attire;
  • representation principles;
  • image style;
  • logo rules;
  • restrictions.

It still cannot decide:

  • which role is being recruited;
  • who the intended applicant is;
  • what the company wants them to feel;
  • whether the message concerns belonging, ambition, stability, technical challenge, or public service;
  • which activity should demonstrate the culture;
  • where the asset will appear.

Those are present communication decisions.

The requester must supply them.

A useful scene seed might be:

An experienced technician guides two new apprentices through the maintenance workshop during their first morning.

The composer can then ask:

  • Is the focus welcome, training, safety, or career progression?
  • Which equipment may be shown publicly?
  • Should the apprentices appear curious, confident, or initially uncertain?
  • Is the image for a careers page or an internal onboarding deck?
  • Where must the headline sit?
  • Should the environment match a real company facility?

The Company Visual Bible provides continuity.
The human supplies the current intention.

This keeps the workflow from becoming a corporate vending machine for branded imagery.

The Visual Register should not make every employee a designer

One possible response to uneven AI use is to demand that every employee learn professional visual direction.

That is neither realistic nor necessary.

Not every person creating an internal presentation image needs to understand lens focal lengths, advanced colour theory, or cinematic lighting.

  • The goal is not to transform the entire workforce into art directors.
  • The goal is to make the minimum necessary decisions visible.

A guided composer can ask plain-language questions:

  • What should this image help the audience understand?
  • Who or what must be shown?
  • What are they doing?
  • Where will the image be used?
  • What should the viewer notice first?
  • What must be accurate?
  • What must not appear?

The company’s Bible can supply the visual rules already established.
Advanced fields remain available to designers and experienced users.

The interface does not need to expose every technical parameter to everyone.
It needs to prevent important decisions from being left entirely implicit.

Different levels of skill should not become different standards of accountability

A company may accept that staff have different creative abilities.

It should not therefore accept that visual responsibility disappears for less experienced users.

Someone who does not know how to direct a complex image can still state:

  • what the image is for;
  • what must be shown;
  • what information is sensitive;
  • which parts must be accurate;
  • who should review it.

They may then hand the brief to a designer or use a governed generator workflow.

The Visual Register can help them create a better request without pretending they are now capable of completing every stage alone.

This distinction matters.

A tool should not turn limited expertise into false confidence.
It should make it easier for people to recognise when they need help.

  • A staff member may produce an acceptable internal illustration independently.
  • A public campaign hero involving products, people, typography, legal claims, and brand reputation may still need a designer.

The existence of an image generator does not erase the difference between those tasks.

Designers inherit the consequences of weak AI processes

When organisations introduce generative tools without a shared workflow, designers often become the repair layer.

They receive outputs with:

  • distorted logos;
  • inaccurate products;
  • impossible anatomy;
  • contradictory lighting;
  • generic expressions;
  • unusable aspect ratios;
  • no space for text;
  • inconsistent brand language;
  • inappropriate symbolism;
  • details that cannot be edited cleanly;
  • uncertain provenance.

The request may still arrive with urgency:

AI already made most of it. Can you just finish it?

The word just hides the work.

The designer may need to:

  • reconstruct the concept;
  • locate the original purpose;
  • regenerate or rebuild the scene;
  • replace product details;
  • correct perspective;
  • recreate logos;
  • extend the composition;
  • change the crop;
  • integrate type;
  • check accessibility;
  • verify that the final result can be published.

In some cases, repairing the generated asset takes longer than beginning properly.

The problem is then framed as designer resistance:

Why is this taking so long when the image was already generated?

The image was generated.
The concept may not have been resolved.
The asset may not have been production-ready.

Those are different things.

Rendering speed is not project completion speed

Generative systems have dramatically reduced the time required to produce visual options.
That does not mean every other stage has disappeared.

A corporate visual may still require:

  1. understanding the communication need;
  2. checking whether generation is appropriate;
  3. protecting confidential information;
  4. selecting the correct company context;
  5. writing the scene;
  6. applying the Visual Bible;
  7. generating or designing options;
  8. reviewing accuracy;
  9. rejecting unsuitable results;
  10. correcting the chosen direction;
  11. adapting it to required formats;
  12. integrating typography;
  13. obtaining approval;
  14. preserving the final asset and decision record.

The rendering step may take seconds.
The project consists of more than rendering.

This is one reason I want The Visual Register to preserve lightweight decision notes.

A manager who sees only the final image cannot easily see why three earlier versions were rejected.

A short record might show:

  • Draft one rejected because the scene resembled a generic technology company and did not reflect the actual workplace.
  • Draft two rejected because protected machinery was visible.
  • Draft three revised to create sufficient headline space and improve accessibility.
  • Final version approved after brand review.

The purpose is not to make the workflow look more elaborate than it was.
It is to make actual judgment visible.

The decision record changes the management conversation

Without a record, discussion often centres on time:

Why did this take two hours if AI generated it?

With a record, the conversation can centre on decisions:

The first concept communicated the wrong audience tone.

The second depicted the product inaccurately.

The third was suitable visually but unusable in the required layout.

This makes design labour legible to people who may not share the designer’s vocabulary.

It also improves future work.

If the same failure recurs, the organisation can ask whether the Company Visual Bible needs a stronger rule.

For example:

Generated laboratories repeatedly appear dark, blue, and speculative.

That may lead to a new Bible entry:

Company laboratories should be depicted as bright, credible working environments based on approved facility references. Avoid blue-neon science-fiction styling and decorative interfaces unrelated to actual work.

The correction becomes organisational knowledge rather than remaining inside one designer’s private frustration.

The company must decide what information may enter a model

Visual consistency is only one part of responsible use.
A company that allows AI also needs to determine what employees may submit.

A prompt may accidentally contain:

  • confidential product details;
  • customer information;
  • internal documents;
  • unreleased campaign material;
  • photographs of private workspaces;
  • employee data;
  • proprietary diagrams;
  • security-sensitive environments.

The Visual Register cannot replace legal, security, or IT policy.
It can, however, make restrictions visible inside the workflow.

A governed Bible or workspace might state:

  • use only approved product descriptions;
  • do not upload photographs containing customer records;
  • exclude confidential machinery;
  • do not enter names of private clients;
  • use the public-facing facility profile;
  • seek review before using employee likenesses;
  • do not upload unpublished technical diagrams to external platforms.

These are not aesthetic rules.
They belong to the same source of truth because they affect what may appear in, or be used to create, the image.

A company that merely says AI is allowed without defining these boundaries has not completed the policy.
It has transferred the risk to individual employees.

Approved platforms do not eliminate process risk

An organisation may approve a particular AI platform after evaluating its security, contractual terms, or data handling.

That is important.

It still does not guarantee that employees will produce appropriate work.

A secure tool can be used with:

  • a poor brief;
  • inaccurate company information;
  • inappropriate visual metaphors;
  • inconsistent branding;
  • unreviewed generated people;
  • inaccessible layouts;
  • misleading product depictions.

Platform approval answers:

May this system be used?

The Visual Register addresses a different question:

What should the organisation ask the system to make, and how will it know whether the result is suitable?

Both layers matter.

  • Security approval without visual method produces safe inconsistency.
  • Visual method without security policy produces organised risk.

The Visual Register belongs to one part of a larger organisational system.
It should not pretend to be the whole system.

The same generator may behave differently across employees

Even when staff use the same approved platform, results can vary because of:

  • different wording;
  • different reference images;
  • different conversation histories;
  • different model settings;
  • different account features;
  • different levels of persistence;
  • different willingness to regenerate;
  • different aesthetic judgment.

One employee may work inside a long thread where the model already understands the company’s environment.
Another begins from an empty session.

One may attach approved references.
Another may rely on the word professional.

One may notice that the company logo has been distorted.
Another may not.

This is why the canonical object should not be the prompt or the generator session.
It should be the human-readable visual brief and the Visual Bible from which it was assembled.

Those can be inspected independently of one model’s behaviour.
They can be handed to another employee, a designer, or another platform.

The output may still vary.

The organisation’s intention becomes more stable.

Standardisation should not mean visual sameness

A Company Visual Bible can be misused.

An organisation may turn every rule into a locked template and produce a monotonous stream of nearly identical images.

That is not the purpose.

The company needs consistency of identity, not sameness of composition.

The Bible may preserve:

  • palette;
  • tone;
  • product accuracy;
  • visual posture;
  • representation principles;
  • environmental language;
  • prohibited elements.

The present brief should still allow variation in:

  • concept;
  • framing;
  • metaphor;
  • action;
  • emotional emphasis;
  • campaign treatment;
  • medium.

A recruitment image, research report cover, internal safety illustration, and product announcement should not all look identical.

They should understand that they belong to the same organisation.
This is the same distinction that applies to artistic worlds.

Visual continuity exists beneath style and format.

A governed system should protect the lineage without flattening every output into one template.

Governance is about authority over persistent decisions

The corporate use of The Visual Register should not be designed as a separate prompt mode.
It is the same composer operating inside a governed workspace.

The difference lies in who controls the inherited information.

A small organisation may need only:

  • one shared Company Visual Bible;
  • approved assets;
  • several locked rules;
  • a draft or approved status;
  • short revision notes.

A larger organisation may eventually require:

  • departmental profiles;
  • campaign layers;
  • role permissions;
  • formal review;
  • audit histories;
  • integration with asset systems.

Those features should be added only when real use requires them.

The foundational idea remains simple:

Governance controls what is inherited, who may change it, and who may approve the result.

It does not replace the creative act.

A company should not confuse governance with surveillance

Making decisions visible does not require tracking every keystroke or turning creative work into administrative theatre.

A useful record may include:

  • the original request;
  • the attached Bible;
  • the resulting brief;
  • key revision reasons;
  • status;
  • approver where necessary.

It does not need to record every experimental sentence, abandoned idea, or private note.

Design involves exploration.

Not every discarded thought belongs in an audit trail.
The purpose is to preserve meaningful project decisions, not to monitor the creator’s every movement.

This matters because a system intended to make labour legible could easily become a system used to measure employees mechanically.

  • A large number of prompt revisions does not necessarily indicate inefficiency.
  • A small number does not necessarily indicate quality.

The record should support communication and responsibility.
It should not become a substitute for judgment.

The manager’s role also changes

Managers cannot responsibly delegate every unresolved visual decision with:

Use AI and make something.

They need to provide enough direction for the request to become meaningful.

They may need to state:

  • the communication objective;
  • the audience;
  • the intended use;
  • the non-negotiable content;
  • the deadline;
  • who approves the result.

They do not need to specify the lens, lighting, and composition unless those are relevant to their expertise.
But they should not expect designers or models to infer the entire organisational purpose from a vague adjective.

The Visual Register can make this visible by requiring a scene seed and purpose before technical choices begin.

A request that cannot answer:

What is the image meant to communicate?

is not ready for generation.

That is not unnecessary friction.

It is unfinished management work becoming visible.

The designer’s role does not disappear

A common corporate assumption is that AI moves visual work from designers to everyone else.
A more accurate description is that it redistributes parts of production.

Non-designers may become able to create:

  • basic internal illustrations;
  • early visual ideas;
  • presentation concepts;
  • rough campaign directions;
  • simple branded assets.

Designers may spend less time manually producing some first drafts.

They may spend more time on:

  • art direction;
  • systems;
  • templates;
  • review;
  • correction;
  • advanced composition;
  • typography;
  • campaign coherence;
  • accessibility;
  • brand governance.

The work changes.

It does not become unnecessary.

In fact, once more employees can generate visuals, the organisation may need stronger design systems rather than fewer designers.

A larger volume of images creates a larger need for judgment.

The Visual Register is not a tool for proving that everyone can design

Its purpose is not to declare:

With enough fields, every employee becomes a designer.

That would repeat the same mistake in a different form.

The tool can help a non-designer produce a clearer brief.

  • It can provide approved visual continuity.
  • It can explain essential choices.
  • It can help them recognise what they do not know.
  • It cannot supply years of visual practice through a form.

The product should therefore support several legitimate outcomes:

  • the user creates a simple asset independently;
  • the user prepares a brief for a designer;
  • the user generates exploratory concepts for review;
  • the user discovers that the task needs specialist help;
  • the team stores a reusable visual decision for future work.

Success is not measured only by whether the user generated the final image themselves.
Sometimes the best result is a request that becomes clear enough to hand off properly.

A shared brief can reduce conflict

Many workplace disputes around AI-assisted design are actually disagreements about unstated expectations.

  • A manager believes the image should feel energetic.
  • A designer believes the brand requires restraint.
  • A requester expects a realistic office.
  • The generator produces speculative science fiction.
  • A reviewer objects to the people shown.

No one agreed on those points before production began.

A shared visual brief places the decisions in front of the group.

It might state:

Purpose

Recruitment image for experienced engineering applicants.

Message

Technical challenge, collaboration, and credible professional growth.

Scene

Three engineers review a physical prototype around a worktable inside the approved testing facility.

Visual posture

Focused and collaborative rather than celebratory or staged.

Inherited continuity

Company facility profile, workwear, product references, palette, and safety rules.

Constraints

No confidential equipment, no exaggerated holograms, no generic glass-boardroom setting, no headline embedded in the generated image.

The team can disagree with the brief before the image is generated.
That is usually cheaper than discovering the disagreement after several rounds of production.

The Visual Bible can reveal that the company has not decided

Building a Company Visual Bible may expose uncomfortable gaps.

The organisation may discover that it does not know:

  • how it wants to depict its workplace;
  • what level of fictionalisation is acceptable;
  • how to represent employees;
  • whether generated people are permitted;
  • what visual tone belongs to different departments;
  • which products may be shown;
  • whether its brand guide is operationally clear;
  • who owns final approval.

That is not a failure of the tool.

It is useful information.

A source of truth cannot be created where the organisation has not made the underlying decisions.

The Register can organise and preserve a policy.
It cannot invent organisational consensus and pretend it already existed.

In some cases, the most valuable output of the process may not be a visual brief.
It may be the recognition that the company needs a real conversation before generating more public imagery.

Corporate culture is visible in images

Images do not merely apply a logo to communication.
They reveal assumptions about the organisation.

A company’s generated visuals may imply:

  • who holds authority;
  • who performs technical work;
  • what professionalism looks like;
  • whether the workplace feels collaborative or hierarchical;
  • whether technology serves people or replaces them;
  • whether customers are treated as individuals or abstractions;
  • whether diversity is lived, staged, or absent;
  • whether the company appears grounded in real work or generic aspiration.

These implications may arise even when no one intended them.
That is another reason visual work cannot be reduced to taste.

A scene contains decisions about bodies, space, attention, work, power, and belonging.
The generator will make those decisions if the organisation does not.

A Company Visual Bible cannot resolve every cultural question.

It can make some of them explicit enough to be reviewed.

The product should remain modest about what it solves

The Visual Register will not solve:

  • all corporate AI policy;
  • security;
  • legal compliance;
  • copyright;
  • employee training;
  • brand strategy;
  • design review;
  • biased model behaviour;
  • poor management;
  • every difference in visual ability.

It should not be marketed as a universal AI-governance platform.

Its contribution is narrower:

  • preserve an organisation’s visual decisions;
  • help requesters state the present need;
  • build a readable brief;
  • distinguish inherited rules from new direction;
  • make key revisions visible;
  • improve handoff among requesters, designers, reviewers, and generators.

That is enough to be useful.

A product does not need to solve the entire institutional problem to fill a meaningful gap inside it.

What a prepared company might do

A company that genuinely wishes to allow responsible AI-assisted visual work may eventually establish several layers.

Tool policy

Which systems may be used and under what data conditions?

Visual source of truth

What should the organisation look like across generated and non-generated media?

Request method

What information must be supplied before visual production begins?

Human review

Which outputs require a designer, brand lead, legal review, or manager approval?

Decision record

What needs to be preserved about the brief, revisions, and final selection?

Learning support

What resources help staff understand scene construction, visual language, and the limits of generators?

The Visual Register sits mainly within the second, third, and part of the fifth layers.

It can support the others.
It should not claim to replace them.

“AI allowed” should eventually become “work defined”

The phrase AI allowed centres the tool.
A more mature organisational approach centres the work.

Instead of asking only:

May employees use an image generator?

the company begins asking:

What kinds of images are staff expected to make?

Which decisions belong to the organisation?

Which decisions belong to the requester?

Which tasks still require a designer?

What information may be shared?

How will the result be reviewed?

What makes the image acceptable?

This shift matters.

The generator becomes one method within a defined process rather than the process itself.

That is where The Visual Register belongs.

  • It does not exist to make the organisation more enthusiastic about AI.
  • It exists to make the organisation more deliberate about visual work conducted with AI.

Allowing AI is the beginning of the question

A company may allow AI because it wants:

  • efficiency;
  • experimentation;
  • reduced production costs;
  • wider staff participation;
  • faster ideation;
  • new creative capabilities.

Those goals are understandable.

But access without method can produce:

  • inconsistent outputs;
  • hidden repair work;
  • weakened brand identity;
  • security mistakes;
  • conflict between managers and designers;
  • false confidence;
  • unclear accountability.

The solution is not to prohibit every use.
Nor is it to assume the tools will naturally teach staff how to use them well.

The organisation needs structures that preserve what it already knows and reveal what still needs to be decided.
That is the working problem behind the corporate side of The Visual Register.

When a company says it allows AI, it has not finished defining the practice.
It has only admitted the tool into the room.

The next question is whether the people in that room share any understanding of what they are trying to make.

© 2026 • MITHAQ PRAXIS • CC BY-NC-ND 4.0 Unless Otherwise Stated.