What we do · 04 · AI Design

AI assistants and AI-made content your customers and teams can trust.

AI assistants, AI-made images and films, and rules to use AI safely.

Services
4
AI models
Tested on your examples
Every output
Approved by a named person

Illustration: the record for one AI image, made by the brand-trained model. Brand score 0.93, above the 0.85 pass mark. Rights cleared, no personal data found. A creative director must still approve it.

What we make

4 AI services, bought alone or together.

Start with any one. All 4 share the same AI models, tests and approval rules.

Over-ear headphones and a microphone in front of an open laptop

01 / 04 · AI assistants, agents

AI Application Design

AI assistants designed to show sources and pass hard questions to a person.

  • Conversation design
  • Agent plans and approvals
  • Voice and camera input

Typical 6–12 weeks to a test version

Explore AI Application Design
An editor works on footage at a monitor in a darkened edit suite

02 / 04 · Images and films

AI Content Studio

Visuals, films and voice-overs made with AI, with usage rights recorded.

  • Creative direction for AI work
  • AI images & film
  • AI voice-overs & audio

Typical 4–8 weeks to a working studio

Explore AI Content Studio
Hands sort a spread of instant photographs on a wooden table

03 / 04 · Brand models

Brand AI Tools

AI image and writing models trained to look and sound like your brand.

  • Training examples you may use
  • Brand-trained image & product models
  • Brand writing models

Typical 6–10 weeks to a trained model

Explore Brand AI Tools
Two people arrange sticky notes on a glass wall during a workshop

04 / 04 · AI rules and trials

AI Strategy & Consulting

AI rules, training and first trials for your marketing and brand teams.

  • AI readiness diagnostic
  • Ranked AI ideas & plan
  • AI tool & licence review

Typical 6-week diagnostic · 12-week trial

Explore AI Strategy & Consulting

How it works

Run a content request through AI tools, checks and approval.

Each job goes to the AI that tested best on your work, then through brand, rights and safety checks.

Studio · Your company Request R-0412

Request

Spring launch for India, the UAE, the UK and Germany. A main image, product text in 4 languages, a voice-over and 8-second films.

Pass mark 0.85 · below it, nothing reaches review.

  1. Text Passed

    Claude · brand voice prompt set

    Best brand-voice score on 120 test prompts

    • Brand
    • Rights
    • Safety
    0.91

    4 × product text · 4 languages

  2. Image Passed

    Brand-trained Flux model · v4

    Only model above the brand pass mark

    • Brand
    • Rights
    • Safety
    0.93

    12 images · 4:5, 1:1, 16:9

  3. Voice Flagged

    ElevenLabs · licensed brand voice

    Consented voice, contract on file

    • Brand
    • Rights
    • Safety
    0.88

    4 × 20 s clips

    Script says “clinically proven”, which is not on the approved claims list.

  4. Video Passed

    Veo · fixed storyboard

    Smoothest motion on your brand tests

    • Brand
    • Rights
    • Safety
    0.86

    8 × 8 s short films

Human approval

1 lane flagged. Voice: script says “clinically proven”, which is not on the approved claims list.

Creative director
Named approver for this brand

Approve stays locked while any lane is flagged. The approver can send a lane back, never skip the check.

Action log

  1. Request read · 4 outputs, 4 markets
  2. Assigned · text, image, voice, video
  3. Auto checks · 11 of 12 passed
  4. Voice lane flagged · claims list
  5. Waiting for a named approver

Illustrative run: an example, not a client result.

Assistant design

AI assistants designed for when the AI is unsure, slow or wrong.

We design for 4 moments in a live assistant. People can check an answer, fix a mistake or stop an action.

Mock: an assistant answer with 2 numbered sources and a medium confidence meter.

01 · Unsure

Sources and confidence shown

Each answer shows its sources and how sure the AI is. People can check what matters.

Mock: an answer being written, showing what it reads, time taken and a stop button.

02 · Slow

Progress shown while it works

Answers appear word by word, so waiting feels shorter. People can stop and keep what is done.

Mock: an answer with a date the person fixed, an undo button and a note that the fix was saved.

03 · Wrong

Fix it in one step

People can edit, undo or ask for a person. Each fix is saved to test the next version.

Mock: an agent’s 3-step plan. The last step, emailing 1,240 people, waits for approval.

04 · Risky

A pause before steps you cannot undo

AI agents act, so they show their plan first. Any payment, message or post waits for a person.

How we use AI

Each AI model picked by tests on your own examples.

We test again whenever a model or your data changes. Every AI task has 4 controls.

  • Your examples first

    Your brand guidelines become test requests and scoring rules for every model and update.

  • Automatic checks

    We check every output for brand, rights, claims and personal data before anyone sees it.

  • A named person approves

    Approval is a fixed step that nobody can skip. The approver sees what was flagged and why.

  • A record for every output

    We store the model, version, instructions, inputs, scores, checks and approver with each file.

A hand holds two prints over a table of photographs laid out to choose from
The pickscored side by side, not by habit
Image lane · model comparison Brand match on 200 test requests
  • Brand-trained Flux model 0.93
  • Adobe Firefly custom model 0.87
  • SDXL with a brand add-on 0.84
  • Base model, no training 0.61
Passes the markBelow the markMark 0.85
Scored by
Auto scorer + 3 trained reviewers
Re-run
Each time a model or data changes
Decision
Brand-trained model · others kept as backup

Illustrative scores. Yours come from tests on your own brand files.

Brand AI Tools compared

2 Brand AI Tools: one makes files, the other checks them.

Brand Design writes your brand rules and checks. AI Design trains AI models on your brand files. Busy teams use both, joined at the brand checks.

Brand DesignBrand guidelines

Logo, colours, voice and rules.

AI Design · Brand AI ToolsBrand-trained models

Trained on your brand, scored against its rules.

Brand Design · Brand AI ToolsBrand checks

Every file checked, from any source.

A named personApproved file

Signed, logged, published.

QuestionAI Design · Brand AI ToolsBrand Design · Brand AI Tools
What it isImage, product and writing models trained on files you have rights to.Rules, checks and templates so all work matches your brand.
What it makesNew files that already look and sound like your brand.A verdict on whether any file matches your brand, and why.
Built withFlux, Adobe Firefly, ComfyUI and open models, trained and scored.Your brand guidelines, turned into rules and an AI brand checker.
You ownThe training data, the tests and the trained model.The rulebook, the templates and the checking steps.
Start here whenImages miss your brand, are slow or rely on a few people.Work from many teams and tools misses your brand.
Brand-trained models Brand checking tools

How a project runs

5 stages from request to handover.

A typical project takes 8 to 12 weeks. We always agree how to score success before the AI makes anything.

  1. Weeks 1–2

    Define

    We agree who uses it and what it may see and do. Limits and test examples come before design.

    • Scope & risk list
    • Test examples v1
  2. Weeks 2–6

    Build

    We build with a live AI model from week 1. Static pictures cannot show how AI answers feel.

    • Working demo
    • Model shortlist
  3. Weeks 4–8

    Test

    We score each model on your examples and test it with your users, including hard cases it gets wrong.

    • Scorecard
    • User test findings
  4. Weeks 6–10

    Set checks

    We set up automatic checks, approvers and a record of every action. We note which AI laws apply.

    • Check settings
    • Approval steps
  5. Weeks 8–12

    Hand over

    Your team takes over. You get the plans, AI instructions, trained models, tests and a how-to for model changes.

    • Specification
    • How-to & training

Tools we use

Our AI tools, with models you can swap.

We group them in 4 layers. When we swap a model, its instructions, tests and checks stay the same.

L1

AI models

Chosen for each task.

  • OpenAI
  • Anthropic
  • Google Gemini
  • Mistral AI
  • Meta Llama
  • Hugging Face

L2

Creating & training

Image, film and voice, and models we train.

  • PyTorch
  • Replicate
  • Modal
  • vLLM
  • ONNX
  • NVIDIA
  • MLflow
  • Flux
  • Adobe Firefly
  • ComfyUI
  • Runway
  • Veo
  • ElevenLabs

L3

Build & automation

AI agents, automation and the app around the model.

  • LangGraph
  • LlamaIndex
  • Python
  • TypeScript
  • React
  • Next.js
  • Playwright
  • Vercel
  • Cloudflare
  • n8n

L4

Design & content tools

Where the work is designed, stored and measured.

  • Figma
  • Storybook
  • Contentful
  • Sanity
  • Notion
  • Miro
  • PostHog
  • GitHub

Standards we build to

Safe AI built into the work, with records to show it.

We follow these laws and standards. They shape our tests, checks and records. We hold no certificates for them.

  • EU AI ActArtificial Intelligence Act (EU) 2024/1689

    We label AI content and chatbots as the law asks, and record each system’s risk level.

  • ISO/IEC 42001:2023Artificial intelligence management systems

    Our AI management follows its structure: clear roles, impact checks and a list of every model in use.

  • NIST AI RMF 1.0AI Risk Management Framework

    Its 4 steps (govern, map, measure, manage) shape our risk list and test plan.

  • OWASP Top 10 for LLM ApplicationsSecurity risks in generative AI applications

    We test our checks against the top 10 AI security risks, like hidden instructions.

  • WCAG 2.2 AAWeb Content Accessibility Guidelines

    Level AA, so people with disabilities can use every AI screen, with captions on AI films.

  • GDPRGeneral Data Protection Regulation (EU) 2016/679

    A legal reason and the least data needed for every use, training included.

  • DPDP Act 2023Digital Personal Data Protection Act, 2023

    India’s data law, applied to AI instructions, logs and training data.

  • We apply each standard only where the work involves it.

What you get

Everything we make, documented so your team can run it.

Instructions, tests, trained models, check settings and approval records sit in your accounts from week 1.

01 · AI Apps

AI Application Design

  • Experience plan, task & risk mapDoc · board
  • Screen and flow designFigma
  • Scripts & AI instructionsDoc · test questions
  • Test version on a working AIDemo · code
  • AI screen parts libraryFigma · Storybook
  • User testing reportReport · recordings
  • Design specification & sign-off checksDoc · tickets

02 · Content Studio

AI Content Studio

  • Creative direction & referencesFigma · PDF
  • Finished files by size & marketFiles · library
  • AI model & settings recordSheet · code
  • Rights, licence & consent recordSheet
  • Automated production stepsSet-up · code
  • Approval steps & action recordSet-up · log
  • Studio how-to & trainingDoc · sessions

03 · Brand Models

Brand AI Tools

  • Rights-checked training & test examplesData · rights list
  • Trained model files or add-onsFiles · your accounts
  • Model notes & training recordsDoc
  • Brand-match tests & reportTests · report
  • Automatic check & misuse test resultsTests · report
  • Model link & version listLink · version list
  • Output log & approval stepsDashboard · steps

04 · AI Strategy

AI Strategy & Consulting

  • Process map & today’s numbersBoard · sheet
  • Ranked AI ideas & AI planSlides · sheet
  • AI tool & licence listSheet
  • AI use policy, rights & disclosureDocument
  • Trial plan, results & decision noteDoc · dashboard
  • Team roles & approvalsDoc · chart
  • Training plan & how-to guidesSessions · docs

What changes

6 measures agreed before the work starts.

We record your starting point in the first weeks. These figures are typical, not client results. Targets are set per project.

01

Tasks completed

Share of test tasks people finish, including ones where the AI is wrong.

  • Before58%
  • Target85%+

Measured by Watched user tests, before and after

02

Brand match first time

Share of AI files that pass the brand score with no manual edit.

  • Before40%
  • Target80%+

Measured by Scorer + reviewers on every output

03

Request to approval

Time from request to signed approval.

  • Before10 days
  • Target3 days

Measured by Times in the action log

04

Cost per approved file

AI costs plus review time for each file released.

  • BeforeIndex 100
  • TargetIndex 45

Measured by AI bills + logged review time

05

Use after 90 days

People using the new way of working each week, 3 months after the trial.

  • Before20%
  • Target70%+

Measured by Weekly users ÷ paid licences

06

Edits needed

Edits a person makes before accepting an AI answer or file.

  • Before5 edits
  • Target2 edits

Measured by Edits per accepted output

Offers & plans

AI Design services, with the scope and timing of each.

Offers from all 4 AI Design services, grouped by what you need done. Pick one to reach the team that does it.

Categories
04
Offers
20
Plans
06
Not sure what you need? Tell us the problem

How to buy

  1. 01Pick offers. Ask about one, or add several to one enquiry.
  2. 02Choose a plan. A small job, a fixed project or ongoing help.
  3. 03Send your enquiry. We reply within one working day.

Browse by category

Timelines are typical. We write the scope before we quote.

01Design & build

5 offers
Typical timeline: 6–10 weeks

AI assistant design

We design what your assistant says, asks and does when it is unsure or wrong.

What’s included

  • Scripts and AI instructions
  • Screens for slow and unsure answers
  • Sources, fixes and human hand-off
  • Live AI test version
  • Assistants
  • Conversations
  • Live demo
  • OpenAI

Best forProduct and marketing teams giving customers or staff an AI assistant.

Explore AI Application Design
Typical timeline: 8–12 weeks

AI agent design

AI agents do tasks for people. Ours show their plan first, with stop and undo.

What’s included

  • Plan, approve and undo steps
  • Permissions and human hand-off
  • Progress, pause and hand-over screens
  • Working agent test version
  • Agents
  • Human approval
  • Undo

Best forTeams building a product that does tasks, not only answers questions.

Explore AI Application Design
Typical timeline: 6–12 weeks

Voice and camera design

AI that takes speech, photos or documents, built for loud rooms, poor light and bad scans.

What’s included

  • Design for speech, photos and scans
  • Confirm, interrupt and fix steps
  • Typed option for every spoken step
  • Accessibility review to WCAG 2.2 AA
  • Voice
  • Camera
  • Documents
  • OpenAI

Best forProducts used hands-free, on the move or without typing.

Explore AI Application Design
Typical timeline: 2 weeks

AI concepts in 2 weeks

One broad AI idea becomes 3 concepts, each shown as a working demo.

What’s included

  • Idea session with your team
  • 3 concepts designed and demoed
  • Effort and risk for each
  • A recommendation with next steps
  • First step
  • Sprint
  • Demos
  • OpenAI

Best forLeaders with an AI idea who want to see it work first.

Explore AI Application Design
Typical timeline: 4–8 weeks

Working test version

A live AI demo with your own content, to test with users or show investors.

What’s included

  • Your data or a safe sample
  • Document search or tool links
  • Tests that cover hard cases
  • Test sessions with your users
  • Working demo
  • Testable
  • Your data
  • LlamaIndex

Best forTeams that need proof before paying for a full build.

Explore AI Application Design

02Create & produce

5 offers
Typical timeline: 4–8 weeks

Content studio set-up

AI makes files in bulk and your team directs the work. We set up tools, rights and training.

What’s included

  • A tool for each content type
  • Rights, consent and AI labelling rules
  • Production steps with a sign-off
  • How-tos and team training
  • Studio
  • Approval
  • Training

Best forBrand and content teams with more work than they can produce.

Explore AI Content Studio
Typical timeline: 3–8 weeks

AI films & animation

Short films and animation made with tools like Runway and Veo, then edited and finished.

What’s included

  • Creative direction and reference
  • Best AI tool for each shot
  • Colour, sound and finish
  • Sizes for every channel
  • Runway · Veo
  • Social
  • Finished

Best forBrands that need video more often than they can shoot it.

Explore AI Content Studio
Typical timeline: 2–6 weeks

AI voice-overs & audio

Voice-overs in many languages from one recording. Every voice has written consent.

What’s included

  • Consent and licence for each voice
  • Voice-overs in your languages
  • Native-speaker review before release
  • AI labels and origin records
  • Voice-over
  • Languages
  • Consented
  • OpenAI

Best forBrands publishing video or audio in several languages.

Explore AI Content Studio
Typical timeline: 6–12 weeks

Personalised ads & emails

Ads and emails that change by customer group, market or product, built only from approved parts.

What’s included

  • Parts and versions designed
  • Build rules and default versions
  • Uses your product or customer data
  • Review before anything goes out
  • Versions
  • Your data
  • Approved parts

Best forMarketing teams reaching many customer groups on one budget.

Explore AI Content Studio
Typical timeline: 4–10 weeks

Product & campaign images

Product photos, lifestyle scenes and campaign versions made from your product files, in every size.

What’s included

  • Product photos or a file check
  • Scenes and versions made
  • Sizes and market versions
  • Rights record for every input
  • Product shots
  • Versions
  • Online shops

Best forRetail and consumer brands with large catalogues and short seasons.

Explore AI Content Studio

03Train & test

5 offers
Typical timeline: 6–10 weeks

Brand image model

An image model trained on files you have rights to, so its images match your brand.

What’s included

  • Training files chosen and cleared
  • Uses Flux, Firefly or open models
  • Brand match scored before release
  • Model placed in your accounts
  • Trained
  • Scored
  • Yours

Best forBrands that need more images than photo shoots can supply.

Explore Brand AI Tools
Typical timeline: 4–8 weeks

Brand writing model

A writing assistant that follows your brand voice and banned words, in your team’s tools.

What’s included

  • Your tone and words as rules
  • Set instructions or a trained model
  • Scored against the voice rules
  • Linked to your writing tools
  • Brand voice
  • Writing
  • In your tools
  • OpenAI

Best forTeams writing a lot across markets, channels and agencies.

Explore Brand AI Tools
Typical timeline: 3–5 weeks

Brand match testing

A fixed set of tests that shows whether each model version matches your brand.

What’s included

  • Brand scoring rules written with you
  • Separate test examples
  • Auto scores matched to human reviewers
  • Score report per model version
  • Testing
  • Version checks
  • Per version

Best forBrands making lots of AI content with no way to judge it.

Explore Brand AI Tools
Typical timeline: 4–8 weeks

Model hosting & versions

Your trained models, linked to your apps and design tools, with versions, cost and speed tracked.

What’s included

  • A secure link to the model
  • Model and version list
  • Limits on cost, use and speed
  • Output records and undo
  • Linked
  • Versions
  • Cost limits

Best forTeams with a trained model only one expert can run.

Explore Brand AI Tools
Typical timeline: 2–5 weeks

Training data you may use

Your files made into training data, each item checked for licence and consent.

What’s included

  • Files collected, chosen and labelled
  • Licence and consent check per item
  • Bad examples and separate test files
  • Data versions and written records
  • Data
  • Rights-checked
  • Versions

Best forAny brand about to train a model on its own files.

Explore Brand AI Tools

04Plan & adopt

5 offers
Typical timeline: 3–4 weeks

AI readiness diagnostic

We map how your team makes content and where time, money and rework go.

What’s included

  • Walk-throughs with each team
  • Time, cost and rework today
  • Tool, licence and skills review
  • Ranked shortlist of where to start
  • First step
  • Costs today
  • Check-up

Best forMarketing and brand heads who need an AI plan they can defend.

Explore AI Strategy & Consulting
Typical timeline: 8–12 weeks

Measured AI trial

One team uses AI for one task, measured against today. Then you decide: go or stop.

What’s included

  • Success measure agreed first
  • Starting numbers taken first
  • Checks and policy from day 1
  • Decision and roll-out plan
  • Trial
  • Measured
  • Go / no-go
  • OpenAI

Best forCompanies with many AI tests and no proof yet.

Explore AI Strategy & Consulting
Typical timeline: 2–4 weeks

AI tool & licence review

We check every AI tool you pay for and advise which to keep, merge or cancel.

What’s included

  • Tools and licences, approved or not
  • Licence, data and legal-cover review
  • Where tools overlap and can merge
  • Advice with steps to switch
  • Licences
  • Unapproved AI
  • Cost

Best forTeams paying for more AI tools than anyone can list.

Explore AI Strategy & Consulting
Typical timeline: 3–6 weeks

AI use policy, rights & disclosure

Rules on what to make with AI, what to label, what data goes in and who approves.

What’s included

  • AI use rules for creative work
  • Rights, consent and origin records
  • Labelling rules per channel and market
  • Approval steps and named owners
  • Policy
  • AI labels
  • EU AI Act

Best forBrands already using AI tools with no written rules.

Explore AI Strategy & Consulting
Typical timeline: 4–8 weeks, then quarterly

Training for creative teams

Hands-on training by role, with how-to guides and use tracked for 3 months.

What’s included

  • Sessions by role, on your projects
  • How-tos and saved instructions per role
  • Team champions and a help channel
  • Use and quality tracked 3 months
  • Training
  • How-to guides
  • Daily use
  • Slack

Best forTeams given AI tools but no guidance.

Explore AI Strategy & Consulting

Your enquiry

Tick “Add to enquiry” on any offer, choose a plan, then continue. Or enquire about one offer.

How we work with you

Ways to start, from a question to a quote.

Ask a quick question, send your project details or request a quote. The person who will lead the work reads each one in full. Services in your enquiry go with it.

Or book a 30-minute call

What are you sending?

  1. 01

    About 2 minutes3 short answers

    For a first chat, a press request or anything not ready to plan yet.

    You get A reply from a named lead

  2. 02Recommended

    About 8 minutes5 short steps

    Goals, who it is for, budget and timing, so our first reply can outline the work.

    You get Options and prices after one call

  3. 03

    About 15 minutesYour documents attached

    Your documents, deadlines and the purchasing and security rules to meet.

    You get Confirmation and a named lead

How it is priced

Each plan shows how it is priced. Work starts once we agree a written scope and price.

  • Typical length
    1–3 weeks
    Pricing
    Fixed fee
  • Typical length
    4–12 weeks
    Pricing
    Fixed price
  • Typical length
    3–9 months
    Pricing
    Fixed price per stage
  • Typical length
    Ongoing · 6-month minimum
    Pricing
    Monthly fee
  • Typical length
    6–18 months
    Pricing
    Priced per written scope
  • Typical length
    Ongoing · 3-month minimum
    Pricing
    Paid by time used
Compare what each plan includes
What each plan includes and who it suits
PlanWhat you getBest for
SprintA short, fixed job that answers one clear question.
  • Scope and result agreed before we start
  • A senior lead and the right specialists
  • A weekly review with you
  • A clear answer or a test version
An audit, a brand check or a test version of an app
ProjectOne clear job for one fixed price.
  • A written scope: what you get and how we agree it is done
  • A named project lead and a fixed team
  • A shared plan with checkpoint dates
  • Source files, yours once paid for
A brand identity, a website, a simple app or a set of tools
MilestoneA bigger build in stages you approve and pay for one at a time.
  • Stages, each with its own scope and approval
  • A go or stop review at each stage
  • Re-plan between stages as you learn
  • Pay for each stage you approve
Work too big for one contract, such as a large app
RetainerA monthly team that runs and improves what we built.
  • Team time set aside for you every month
  • Agreed response times for requests and fixes
  • A monthly review and a to-do list
  • Planned improvements, not only fixes
Monthly social media and ads, or care for live sites and apps
EnterpriseA large, long project with its own team and agreed response times.
  • A project director and a decision group
  • A team for each part of the work
  • Response times, reports and a list of risks
  • Security, legal and purchasing checks in the plan
Large companies making changes across markets, brands or business units
SquadOur specialists join your team and work in your tools.
  • Named specialists matched to your plan
  • Works in your own tools and meetings
  • Grow or shrink the team each month
  • We teach your team from week one
Teams with a clear plan that need senior people fast

Questions

What buyers ask about AI Design.

Answers on AI models, ownership, rights, the law and where your data is kept. Ask us anything else.

Book a call
Two people talk across a small table over cups of coffee
The first callyour questions before ours
Which AI models do you use?

The ones that test best on your own examples. For text: Gemini, OpenAI, Anthropic, Mistral or open models. For images, film and voice: Flux, Adobe Firefly, Runway, Veo and ElevenLabs. Each task has a backup.

Who owns the instructions, tests and trained models?

You do, including any brand model we train. What we make for you is yours once it is paid for. Tools we already had stay ours, and you get a free, permanent licence to use them.

How is AI Design different from AI in Technology & Intelligence?

AI Design shapes what people see and use: conversations, content and brand models. Technology & Intelligence builds the software behind them. One project can use both.

How do the 2 Brand AI Tools differ?

Ours makes files in your brand style. Brand Design’s version checks any file against your guidelines. The two meet at the brand checks.

How do you handle rights in AI content?

We clear every input first: licensed or owned images and consented voices, never copies from the web. Each file records its AI model, inputs and approver. If a tool limits business use, we say so.

Does the EU AI Act apply to our marketing content?

Yes, its labelling rules can. Chatbots must say they are AI, and AI-made images, audio and video need a label. We check each system’s risk at the start and build the labels in.

Can our data stay in our own cloud or region?

Yes. We use business plans that never train on your data. If files must stay in your network, we run open models in your own cloud, including in India.

Tell us what you need built.

You will talk to the lead who would run your work.

Book a call

Three ways to start

We always agree a written scope and price before work begins.

Choose one of the three ways above