The anti-AI AI company
AI isn't always the answer. We look at how your team actually works, find what needs fixing, then see where AI helps.
Updated 22 September 2026
Dear Founder,
Are you really more productive with AI?
You were told AI would change everything.
So you got ChatGPT.
Then Claude.
Then you learned about prompts.
Better prompts. Prompt libraries. System prompts.
Then it was workflows.
Then agents.
Then skills.
Then context.
Then MCPs.
Then Obsidian.
Then OpenClaw.
Every week, there was a new thing you apparently needed to understand.
So you sent the links to your team.
You bought everyone accounts.
You ran the workshop.
You told them to experiment.
Some did.
Most went back to doing things exactly the way they’d always done them.
So you pushed harder.
“Why aren’t we using AI for this?”
“Can we automate that?”
“Surely AI can do this now?”
You asked people to build skills.
Someone built an agent.
Someone else built a workflow.
A few people became obsessed.
A few quietly resisted.
And half the company started wondering whether the thing they learned last month had already been replaced by something they were supposed to learn this month.
Meanwhile, LinkedIn kept telling you that if you weren’t rebuilding your entire company around AI, you were already behind.
More agents.
More tools.
More acronyms.
More noise.
And somewhere underneath all of it is a much simpler question:
Is any of this actually making the business better?
Because maybe you don’t need another AI tool.
Maybe your team doesn’t need another workshop.
Maybe the answer isn’t another agent.
Maybe we started with the wrong question.
Instead of:
“Where can we use AI?”
Start with:
What needs to get better?
Something is too slow.
Something costs too much.
Something takes six people that should take two.
Something your best person does still lives entirely inside their head.
Something customers hate waiting for.
Something your team does every Monday that nobody has stopped to question.
That’s the work.
Find it.
Understand it.
Fix it.
And if AI is the best way to do that, use the hell out of it.
If it isn’t, don’t.
We’re not here to help you do more AI.
We’re here to help you build a better business.
AI just happens to be an extraordinary tool for doing it.
We’ll help you start in three ways.
Audit and identify your weakest processes
Sprint
One day inside your business, then six turning what we saw into your roadmap. You leave knowing what is holding you back, and what to fix first. Everything starts here.
See how it worksUpskill your team
Workshops
Focused training for people who already have the tools but are not getting the most out of them.
Attend a workshopHire us to build it for you
Deployment
We deploy agentic AI systems into your business, either scoped to the Sprint's roadmap, or you tell us exactly what you need.
Tell us what you needStop wasting time and money. Start working faster and better.
Skip the guesswork, the platform hopping and the quiet slide back to the old way of working.
Instead, find out exactly what is blocking your team, fix it in the order that matters, and build systems people keep using.
Not another subscription nobody opens.
7 days and you will know exactly how to move forward with AI.
Most of what limits a team is already in plain sight. We uncover that in 7 days.
You leave knowing your level, your gaps, and the order to fix them in.
We sit with you and your team to see what is really there. Friction, edge cases, and the systems people actually use.
We consolidate what we found, time-map the work, score each dimension, match it to the solution library, sequence the fixes, then apply judgment.
We present the report to you in a working session: your level on each dimension, what is behind each gap, and the fixes in the order they have to happen.
Deploy your first agentic AI system in 4 weeks.
Deployment starts with whatever the roadmap put first, so nothing is built before it is needed. This is the shape a typical first month takes.
Foundations
Your documentation moved to where AI can read it, and the standards written down properly.
First workflow built
Built in your environment, on your stack, with human approval gates before anything goes out.
Your team runs it
We watch them use it, fix what breaks, and hand it over properly. Not a training deck.
The next thing on the roadmap
A second workflow, or the next dimension, in the order the Sprint already set.
Seven working days, and you will know your first move.
See how it works →Blueprint was built by the two people who got burnt first.
Pictured moments after telling a room of founders to f*ck AI. He stands by it.
Robert Lai
Rob is an ex-Googler and the founder of Kaliber, a performance marketing agency that has managed more than USD$500m in ad spend across eight APAC markets.
Over the past 18 months, he spent significant time and his own money figuring out why AI adoption kept failing. Blueprint exists to make sure other businesses don't have to learn those lessons the hard way.
He wrote the audit methodology every Sprint runs on.
Shawn Lee
Shawn spent six years in media strategy at MediaCom and PHD, planning Google's campaigns across APAC, before building MoneySmart's brand team from scratch.
As COO of Kaliber, he rebuilt how the agency works, turning the way people work into systems that could scale. If Rob found the problems, then Shawn worked out how to fix them.
Most teams have the tools. Fewer have them connected.
Press the button to see the same team and the same tools, arranged as one system instead of eleven separate subscriptions.
This is not for everyone.
Worth a conversation if
- You have spent money on AI and cannot show what it gave you back
- Your team tried AI, liked it, and quietly went back to the old way
- You want to know the first move before you commit to a big one
- You are ready to commit, and can give us time with the people doing the work, not only the people managing it
Not worth it if
- You want an agent built next week, and none of the work that has to happen first
- Nobody senior can clear a day for the on-site
- You want to satisfy a board with a report rather than change how work is really done
- Your team comprises AI superstars and you think your workflows are perfect
S$8,000 for the Sprint. Everything after is priced from your roadmap.
Here is what you are choosing between.
| Internal experimenting | Hire an AI engineer | AI Readiness Sprint | |
|---|---|---|---|
| Upfront cost | Feels like zero | S$10-15K monthly | S$8,000 fixed |
| Time to results | Unknown | Months | Seven days |
| Ownership | Whoever built it | The engineer | Your team |
| If they leave | Lost capabilities | Lost capabilities | Business as usual |
| If foundations fail | May not discover | Months wasted | Found during the Sprint |
The things people ask us on the first call.
What industries do you work with?
We are fairly industry agnostic, because we focus on functions within companies instead. A skincare brand and an industrial B2B supplier may run nothing alike, but their marketing and sales will contain similarities in the backend and execution.
Are you consultants?
Yes and no. There certainly are similarities to consulting for our AI Readiness Sprint, but what sets us apart is that we actually build and deploy ourselves, while a consultancy hands you a recommendation and leaves.
Do you have a team of engineers?
Yes. Deployment is done in-house, with Shawn leading the build team. Where a project grows in scope or complexity, we bring in senior advisors for the specific technical depth it needs, rather than pretending we hold every specialism on staff.
What happens after the system is deployed?
We train your team to run and maintain it, and the documentation is written for them rather than for us.
After that it is your call: bring us back ad hoc when something needs changing, book us against specific milestones on the roadmap, or keep us on a fractional basis. Nothing is built so that you have to keep paying us to keep it running.
Is my data safe?
Nothing leaves your systems without you agreeing first. We scope data handling in writing before any access is granted.
What does it cost to run once it is built?
It depends on what gets built, so the roadmap sizes it before you commit to anything. For most marketing systems it is model usage plus tools you may already pay for, and we set usage limits so the bill cannot run away.
Why should I trust an agency to tell me about AI?
You should not, on the claim alone. Everything on this page happened to us first, and we will show you any of it running live.
What is an AI readiness assessment?
A structured audit of how your team works today: what is documented, who can actually use AI, how work really moves, and what your tools can do.
Can we just do this ourselves with ChatGPT?
Some of it, yes. What stalls most teams is not the tool: ChatGPT only knows what you paste into it, it stays with the one or two people who enjoy using it, and teams start with what looks impressive instead of what matters. The Sprint tells you which parts you can do yourself.
Should we hire an AI engineer instead?
Not first. An AI engineer costs S$10,000 to S$15,000 a month, and most companies hire one before they know what the role should build. The Sprint tells you whether you need one, and what they should build first.
Will this lock us into one AI vendor?
No. The documentation and processes we build are yours, and they work with whichever AI model you use. If you switch tools, they come with you.
What happens if our readiness comes back low?
You get the gaps in order, with costs attached. Fix them yourself using the plan, or have us work through them with you.
Find out what your first move actually is.
Seven days. One roadmap. No agent gets built until you know why it should exist.
The AI Readiness Sprint
Seven working days. We look at how your team works today, then hand you a roadmap of what to fix first.
Four dimensions decide whether AI works for your team.
Eighteen months of building our own systems taught us that getting AI to work is less about tools, and more about how teams work.
Every wall we hit came back to one of four dimensions. We score all four in seven days, so you know what to fix.
01
Context
The depth of documentation available for AI to reference as context for how your company thinks.
02
People
Who are your champions, who are curious, and who are falling behind.
03
Process
How the work is executed, along with identifying blockers and pain points outside of SOPs.
04
Platform
What tools you currently pay for, and how effectively they are being used.
We turn your gaps and weaknesses into a full development roadmap.
Six sections covering what we found, where you stand on each dimension, and the order to fix things in.
Open any line to see the detail. Every finding carries the evidence behind it.
01Workspace audit
The stakeholders involved, what we inspected, and the evidence behind every finding.
02Insights & gaps
Where your team's time is wasted, what breaks, and the work your team should be doing.
03Level of readiness
Your measured AI readiness level, so you understand what needs to be prioritised and how to move up the ladder.
04Goals & KPIs
We work with you to establish tangible goals and outcomes that should come with correct AI adoption.
05Deployment roadmap
Our Sprint will give you the detailed steps needed with instructions to reach your goals.
06Free AI resources
We'll enable you for the easiest wins so you can get results even if you don't deploy with us.
What you're actually buying.
The language around AI makes this sound more complicated than it is.
We are solving ordinary operational problems. These are the five things you will know at the end.
Where are we, actually?
Your measured level, against the level you claimed on the first call.
Which workflow first?
The priority workflow, and the bottleneck inside it.
What's blocking it?
Every gap named, each with the evidence that found it.
What do we do?
The roadmap: sequenced by prerequisite, sized, with effort and cost estimates attached.
What do we stop paying for?
The what-not-to-do list. It routinely pays for the Sprint by itself.
We start with people first, always.
We start with leadership to understand the AI vision and mandate, then work downwards to daily operators to see how things really operate.
Then we bring both views together to find the gaps and weaknesses, and build a roadmap showing where existing tools need optimisation, where AI helps, and what needs changing.
5 steps to our Sprint.
We deep dive into how your business actually works, assessing your people, processes, tools, and AI opportunities, before turning our findings into a strategic roadmap.
Define your scope
Your setup, the level you think you are at, your target, and your own words for good.
Map out workflows
Your team logs what they actually do: tasks, tools, frequency, ownership, and where AI already sits.
A day on site
The systems actually in use, the friction nobody logs, and the gap between what leadership mandated and what happens on the ground.
Evaluate and sequence
We spend 5 days internally to process the information: we map each task, score each dimension, and put our findings into a prioritised roadmap.
Deployment roadmap
We come back and present our findings in-person: your level, what sits behind each gap, the roadmap, and how this matches your vision.
The 6 levels of AI adoption
We don't grade how many tools you own. We track where context lives, whether it stays current, and whether work can flow through it.
Open any level to see what it actually looks like in a team.
L0Nothing
No AI in the work.
Tools and fixed automation may exist, but nobody uses AI in their work.
L1Chat
AI used like a chatbot.
Prompting may be understood, but every input arrives by hand: pasted, attached, or linked. Nothing persists, so the value leaves when the person does.
L2Agent lite
Encoded once.
System prompts, Projects, and bounded off-the-shelf agents capture understanding inside a container you don't own. Understanding is frozen, so when plans change the AI doesn't know.
L3True agentic
Living context begins.
Agents run in your environment. When reality changes, their understanding changes without re-teaching. Each agent still lives alone, so the next one costs what the first one did.
L4The agent team
One shared substrate.
Specialist agents read from and hand off through a common source of truth. Humans are still the loop: every piece of work begins because a person began it.
L5The function runs itself
The loops initiate.
Agents run mostly autonomously. Humans provide vision, orchestrate, and hold permanent judgment gates rather than carrying every step.
Not sure where you sit? Score your team in four questions, one for each dimension.
We'll help you level up for AI.
Each move up the ladder needs a different kind of work. The roadmap names which one you need, and stops there.
Free resources
Self-served guides that build trust and baseline capability.
Workshop or lite deployment
Done-with-you implementation, or hardened tools in your existing stack.
True agentic deployment
Agents in your environment with living context.
Substrate engagement
Shared context, connectors, and structured data that amortise future agents.
Orchestration design
Autonomous loops with permanent judgment gates.
So how successful is your company's AI adoption?
Enough reading, here's a little taster for you: four questions, one per dimension.
How would you score?
Context. Where do your brand rules, briefs, and campaign standards live?
People. Who on the team gets real work out of AI?
Process. How does a campaign get from brief to live?
Platform. How deep does your AI setup actually go?
Your lowest scored dimension sets your level, because foundations cap the whole function.
Teams typically score themselves one level above where the evidence puts them, and the interesting part is always why.
What it costs to accelerate adoption and deployment.
S$8,000 fixed, for seven working days and the report that comes out of them.
Everything after that is priced from your roadmap, so you see the number before you commit to it.
- Includes intake, working session, standalone report, and briefing
- Excludes training and build work. Those are next steps, priced on the roadmap
- If we don't think you require a custom solution, we'll tell you why and provide you with better ways of working
Asked by search, or asked in the room.
What does an AI readiness assessment actually cover?
Four dimensions: Context, what is written down; People, who can use AI; Process, how work really moves; and Platform, your tools and how deeply you run them.
Do we have to replace our existing agencies, tools, or website?
Usually not. Most gaps are fixed with better documentation and training, not new suppliers. Where something already works, the report says to keep it.
What does it cost to run AI systems after they are built?
It depends on what gets built, so the roadmap sizes it before you commit to anything. For most marketing systems it is model usage plus tools you may already pay for, and we set usage limits so the bill cannot run away.
Can the AI change things without us knowing?
Only within limits you approve. On our own ads system, the agents cannot change a budget by more than 30% or shift more than S$500 of spend in a day. Those limits are written into the system, not left to judgment.
Will this lock us into one AI vendor?
No. The documentation and processes we build are yours, and they work with whichever AI model you use. If you switch tools, they come with you.
Should we hire an AI engineer or run an assessment first?
Not first. An AI engineer costs S$10,000 to S$15,000 a month, and most companies hire one before they know what the role should build. The Sprint tells you whether you need one, and what they should build first.
What happens if our readiness level comes back low?
You get the exact gaps in order, with costs attached. Fix them yourself using the plan, or have us work through them with you.
Why is the assessment not free?
A free audit is a quote wearing a costume: it exists to sell you the build. The fee keeps the verdict honest.
The Sprint is step one. It prices the roadmap. You choose what gets built.
Deployment, workshops, and fractional support all start from the same verdict.
Not sure where to start?
Tell us what you're trying to fix. We'll point you at the right first step.
AI Workshops
Practical workshops for varying adoption levels, so your team can level up how they work with AI.
AI training fails when you apply the same approach to everyone.
People adopt AI at different speeds.
Our workshops focus on specific skills for varying adoption levels, so you can learn what you need, at the level you're ready for.
3 objectives for every workshop.
Understand your gap
Know where you're falling short, what's causing it, and what needs to change.
Solve a real bottleneck
Apply what you've learned to something your team is actually dealing with.
Know what comes next
Understand the next level, and what you need to do to get there.
Open sessions, free to join.
Small rooms, big goals. Each session is focused with clear outcomes you can take back to work.
Seats are limited. Luma registration links below.
Interested in a corporate workshop for you and your team?
We tailor the workshop to your team's current AI adoption, focusing on the skills they need to build next.
Choose the format that fits your goals: leadership alignment, team capability building, or focused sessions for specific roles and skill gaps.
Our founder, Rob, speaking onstage at Ahrefs Evolve 2026.
What actually happens in a workshop?
You bring a real task from your own week. We build it with you in your own accounts, and it runs before you leave.
Is this generic AI training?
No. Generic training creates users for a week. We work on your tasks, in your tools, and write down how to repeat it.
Who should come?
The people who own the work, plus someone who can approve how it runs. Observers use up a seat without leaving with anything.
Do I need to be technical?
No. If you can describe a task you do every week, you can build it in the room. We sit beside you when it breaks.
Why does registration need approval?
Seats are limited because everyone builds. We check the room is the right mix first, so nobody spends ninety minutes watching.
Can you run one just for our team?
Yes. That is the corporate format: your office, your accounts, your team's own work. Length and shape are built around what you need.
How is a workshop different from the Sprint?
A workshop builds capability on one task. The Sprint measures the whole function over seven days and returns a roadmap. Workshops are the move from level one to two.
What does it cost?
The public sessions are free. Corporate sessions are quoted on format and length. Ask, and we will price the one you need.
Workshops prepare the team. The Sprint decides what gets built.
Have a room already?
We'll bring a session to your team or your community.
One method. Every deployment shaped differently.
Each case below is a separate build, in a separate business, with its own architecture. Open any one of them.
What we built, and what it changed.
Agents are named by category and by the job they do. Client names appear once each company agrees to it.
The weekly ads loop, run by agents
A four-person team runs 35 client accounts and over S$500k of media a month.
Read the case study → 02 · A water-purifier brandFive landing pages, under five hours of build
A landing-page method was proved against the client's live pages in a split test and cut the cost of a lead by more than half.
Read the case study → 03 · A US ecommerce mattress brandDelivering 50 ad creatives in a day, not two weeks
A batch of ad creative used to take two weeks and three rounds of revision.
Read the case study →Why are some clients unnamed?
Because we publish with permission, not without it. The architecture and the numbers are real whether or not the name is on it.
Are these numbers real?
Yes. Every figure is counted, dated, and reproducible. Ask on a call and we will run any of them live.
Why does every case look different?
Because the architecture follows the business. Different context, different connections, different approvals. A single template would not have survived any of them.
Will our numbers look like these?
Not on day one. The Sprint tells you which of these gains are reachable for you, and in what order.
When do new cases get published?
As engagements complete and results become measurable. Real names only, and nothing published before there is something to measure.
Ask to see any of these run live.
The weekly ads loop, run by agents
Four people ran 35 client accounts through a weekly loop. Agents now handle the work, leaving one person to make the decision.
The weekly cycle, measured on the agency's own book.
Kaliber Asia
Kaliber Asia is a performance-marketing agency in Singapore, managing paid media for more than 35 client accounts.
It is Blueprint's sister company, and the first place we ran this system. What worked there became the pattern we now build for others.
- Book
- 35+ client accounts on Google Ads and Meta, over S$500k of media a month.
- Team
- Four people run every account.
- Before
- Decks, spreadsheets, and ClickUp for passing work from one person to the next.
- Provenance
- Built inside Kaliber before Blueprint existed. The founders run their own clients through it every week. Blueprint now brings the pattern to other businesses.
Tribal knowledge was the root cause of the problem.
Kaliber's weekly optimisation loop passed through four people. Numbers, reporting, diagnosis, recommendations, and execution all moved through ClickUp.
The real problem was not the handoffs. Each account depended on knowledge held by the person running it, so others had to check their work.
Capacity, consistency, and covering for leave were all one problem: the standard lived in people's heads, not where the system could use it.
We engineered the loop into sets of agents, then gated it with our human decision at the end. The standards became written context the system could work from.
Our knowledge deployed into context and agent skills. Our judgement gates the actions taken.
The diagram is one client account across one week. Each box does one job. Four parts are worth looking at closely.
Read before every step
Every agent reads the account plan and history before it starts. The system begins with the same context, whoever is running the account.
The only human step
Only one step needs a person. Agents prepare the recommendation, then the operator amends, approves, defers, or rejects it. Nothing runs unsigned.
A written rule beats the system's own judgement
Each account carries standing client instructions: never pause this market, never drop below this budget. Those rules override whatever the agents conclude.
What makes it improve
Every approved change is logged with who signed it and when to review it. Next week's work starts with the last week's record.
We went in for productivity. We came out with continuity.
Every morning, agents check all 35 accounts and flag only the ones that need attention. The operator spends about five minutes checking the book.
At the start of each week, agents produce the report, diagnosis, and recommendation for every account. Each account sends one message when its proposal is ready.
The operator then spends up to 60 minutes per account reading, adjusting, and approving the proposal. Approval releases the change.
Time that used to go into reporting and handoffs now goes into clients, strategy, and new work. The same four people can cover more ground.
An agent is only as good as what it reads
The first weeks were spent cleaning up the documents. Old campaigns, changed goals, and conflicting decks had to be resolved before the agents could trust the context.
That work is done once. The system now flags documents that no longer match the plan, instead of quietly making decisions from bad information.
What we can measure, and what we estimate.
| Before | After | |
|---|---|---|
| Human time per account, per weekly cycle | About 12 hours: 4 on the report, 4 on reviews, 4 on making the changes | 60 minutes at most, spent on the decision |
| Time to answer a client question about their account | Within 4 hours | Within 2 hours |
| Changes logged with who approved them and when to review | About 1 in 3, loosely, inside ClickUp tasks | Every change, in one log |
| One-off document pass before the loop could run | None | About 4 hours per account |
- About 11 hours returned per account, per weekly cycle.
- Around 140 hours spent once to clean up documents across 35 accounts.
- Finding a past change went from searching ClickUp to one lookup in the change log.
The same pattern now runs Kaliber's ad creative, websites, landing pages, and search work.
Each started with the same principle: write down the standard, let the system run the process, and keep human judgement at the decision point.
Find out what your first move actually is.
Five landing pages, under five hours of build
A client asked if we could help them create landing pages: our system shipped five landing pages in less than 5 hours, also cutting CPL by 58%.
Same campaigns, same keywords, same bids. Only the page changed.
A leading water-purifier brand
A premium water-purifier brand in Singapore. Its website captures enquiries, which are then followed up and converted by a sales team.
That makes the website a critical part of the acquisition journey. Paid traffic only creates value when those clicks turn into qualified leads.
- The model
- The lead form is the handover from marketing to sales. The page's only job is to earn it.
- Scope
- Lead generation. Google Ads search campaigns, and the pages the ads land on.
- Before
- Six search campaigns sending paid clicks to the site's existing product pages.
Their ads were fine. But the landing pages weren't built to convert the traffic sent.
Paid search clicks landed on product pages built for browsing, not for the promise made in the ad. And a lead cost S$62.
The bigger issue was knowing which pages to build and what they needed to say. Luckily, search and ad data showed us what people wanted and what messages drove clicks.
Instead of guessing, we tested the method against the live pages first. The results gave us a clear benchmark for the new pages.
We built four pages, tested them against the live pages, and used the result to build the next five from a proven method.
Seven stages, each with a defined input, a defined output, and a check.
The speed did not come from writing faster. It came from making the build repeatable, so pages could be produced in parallel and checked before they reached the client.
Which pages should exist
Search data showed us where people were looking, helping us prioritise the pages most likely to turn that demand into leads.
Competitors are read from the results
The search results showed which competitors were winning the click. We used that evidence to identify gaps the new pages needed to close.
Build in parallel, then check
Four pages were built at once. A mechanical check caught broken links, unsupported claims, missing images, and inconsistencies before the client saw them.
The ad and page match
Every live ad promise was mapped to a section on the page. The visitor gets the same message after the click, not a generic product page.
Five pages in two weeks, and most of that was waiting on the client, by design.
The build itself took 4 hours 49 minutes. The engagement lasted two weeks because client review, not production, was the slower part.
The review tool showed the live page beside focused questions. Shared decisions were answered once, and the client submitted one structured set of corrections.
The final bundle included the pages, assets, deployment instructions, and checks. The client's developer deployed all five pages unaided.
What we can measure, and where it went next.
| Existing pages | New pages | Change | |
|---|---|---|---|
| Spend | S$2,179 | S$1,544 | n/a |
| Clicks | 804 | 580 | n/a |
| Leads | 35.0 | 59.3 | +69% |
| Conversion rate | 4.35% | 10.23% | 2.4× |
| Cost per lead | S$62 | S$26 | −58% |
- More leads came from fewer clicks and less spend. The test stopped on day eight because every key signal had moved in the same direction.
- Five pages and the review tool took 4 hours 49 minutes to build, measured two ways.
- 47 client corrections were applied across three passes.
- All five pages were deployed on the client's domain by their developer, unaided, first time.
The method was written down as five reusable stages and two operator playbooks.
A month later, the same review tool was rebuilt in a day for a product brochure on the same account.
Find out what your first move actually is.
Delivering 50 ad creatives in a day, not two weeks
Ad creative took two weeks and three rounds of revision. Now the client approves the ingredients once, and agents produce 50 assets in a day.
The client approves once. The batch ships in a day.
An ecommerce mattress brand in the US
An ecommerce mattress brand selling directly to consumers in the US, with paid social on Meta as its main acquisition channel.
Its performance agency manages the account, with account managers running the creative process from brief through approval and production.
- Working through
- Its performance agency, a US firm with several hundred clients and no creative team of its own. The agency's account managers ran the process. The brand's stakeholders were the sign-off.
- Scope
- Ad creative for paid social. One campaign batch of about 50 assets across three placements.
- Before
- Briefs by email, assets designed by hand, revisions until the stakeholders were satisfied.
Approval was happening at the wrong end of the workflow.
Creative took about two weeks because approval happened after the work was designed. Each round was another attempt to discover what the client wanted.
The team was not slow. The workflow made the client react to finished work, so every revision came after the expensive part.
We moved approval to the start: agree the colour, copy, and layouts first, then let the system produce the assets.
The client approves the ingredients. Agents make the ads.
The diagram is one campaign, from input to finished assets. Six parts explain why the process works.
What every client starts from
Kali Jam Board curates layouts and copy patterns, all proven from Kaliber's own work and across the winning ads of the client's category.
The only input that changes
The client's brand, competitors, and past performance shape the Board. It combines what previously worked with the brand's own guidelines and preferences.
Three questions in a fixed order
The client settles colour, tone, and layouts first. Each choice includes a reason, turning preference into a rule the production agent can follow.
The output is a packet
The Board returns the approved palette, copy, layouts, do's and don'ts, and reasons. That packet becomes the production brief.
Production is an agent
The production agent reads the packet and campaign brief, then builds the assets in Figma. A strategist makes the final pass.
What makes it compound
Each campaign feeds its results into the next one. The system learns what worked for that brand, while the approval step stays fixed.
From a two-week production cycle to eight hours.
The first campaign takes about a week. Setup and client approval take most of the calendar time; production itself takes about four hours.
After approval, the agent produces 50 assets across placements. The strategist checks the set in Figma before anything is used.
The client is involved before production, not after it. They make the key choices once, rather than sending finished work back through three rounds.
On later campaigns, setup and re-approval largely disappear. The strategist briefs the agent, it produces, and the strategist finishes the set.
Where it pays
The Board already exists, so later campaigns take about five hours of Kaliber time, with no new client approval.
What we can measure, and what we estimate.
- A second campaign for the same brand took about 5 hours of Kaliber time, with no client re-approval.
- The overhead of carrying a two-week revision thread is gone with the rounds. Nobody clocked it before, so it is not in the table.
The same pattern is now being extended beyond ads into written copy.
The agency's next step is to run its own Boards for its own clients.
Find out which workflow yours is.
Let's talk about where your team is currently at.
Tell us where your marketing is and what you've already tried.
2 minutes to start your team's AI transformation.
Be as honest and detailed as you can, it gets you a more useful first conversation.
If you are unsure about a field, leave it blank.
The fastest route. One of us reads it, not an inbox rota.
Where the working notes and the method get published.
Operating across APAC: SG, MY, ID, PH, TH, VN, HK, AU.
Not ready to talk to anyone yet?
Score your team on the four dimensions first. Four questions, no email required.