The Spark Report – AI in agencies: Where's the value? Autumn/Winter 2026/27
Not seeing a return on AI? You've sped up the old model. Only 9% of agencies have redesigned how the work gets done. Are you one of them?
Key findings
What's changed in six months
Six months on from our Spring report, here's what has moved – and what hasn't.
Swipe for all six findings
Everyone uses AI.
of agency staff use AI at work – and 91% of them save time.
The time saved keeps growing.
of AI users save five hours or more a week – up from 5% in 2025.
Saved time leaks.
use the time saved for new kinds of work.
Rules exist. Nobody checks.
of agencies don't actively check how AI is used.
AI is bolted on.
of agencies have redesigned a workflow end to end.
Almost nobody has changed how they make money.
have built a new AI-enabled revenue model.
If you take away anything from our findings, take away this
- 1Redesign the work, don’t just speed it up.
Most agencies are using AI to do the same work faster. But only 9% have redesigned a workflow from start to finish. The bigger opportunity is to rethink how work gets done, what becomes possible and what new services agencies can offer.
- 2Faster should cost more.
Hourly pricing turns every efficiency into a discount. Price the outcome and productise what repeats.
- 3Skills are the next advantage.
Skills and culture are the hardest part to move, and they lag in larger agencies. Agencies that build shared skills – and a new way for juniors to learn – will get more from AI.
The big picture
Everyone's faster. Few agencies are better off.
We found that AI use is almost universal, but generating new revenue from it is rare.
How many use AI – and how many turn it into new value
Five signposts from this report, %
See the data
| Staff using AI | 90% |
| Users saving time | 91% |
| Saved time → new work | 9% |
| Redesigned workflows | 9% |
| New AI revenue model | 6% |
Sources: Spark AI pre-programme staff survey (first three) and Spark AI Maturity Score (last two), Autumn/Winter 2026/27. Bases differ – read as signposts, not a funnel.
The Spark AI Maturity Model®
The average agency scores 2.2 out of 4
Scored across Vision & Strategy, People, Skills & Culture, Data & Tools, and Governance & Accountability. Source: Spark AI Maturity Score, Autumn/Winter 2026/27, n=108.
Swipe to see all four stages
Increasing maturity →
Stage 1Experimentation
- Vision & Strategy
- No AI strategy exists
- People, Skills & Culture
- Self-taught, uneven skills
- Data & Tools
- Personal tools, no data structure
- Governance & Accountability
- No policy, down to the individual
How we help you get there
AI FoundationsHalf-day workshop→Stage 2Adoption
- Vision & Strategy
- Goals agreed, not written down
- People, Skills & Culture
- Shared training, sharing becomes routine
- Data & Tools
- Company accounts, single source of truth
- Governance & Accountability
- Policy exists, adherence unchecked
How we help you get there
AI Accelerator10-week programme→Stage 3Optimisation
- Vision & Strategy
- Documented strategy, commercial model shifting
- People, Skills & Culture
- Role-specific development, time protected
- Data & Tools
- Configured assistants, structured knowledge
- Governance & Accountability
- Policy checked, formalised in client contracts
Stage 4Transformation
- Vision & Strategy
- AI is a competitive differentiator
- People, Skills & Culture
- AI-native, continuous by default
- Data & Tools
- Connected systems, agency second brain
- Governance & Accountability
- Proactive governance, a commercial asset
How we help you get there
Spark FrontierMonthly momentum for your AI programme→
Where agencies sit today
On every dimension, the average agency is in early Stage 2. Governance is furthest ahead, at 2.3.
Average score on each dimension
Spark AI Maturity Score, 1–4
See the data
| People, Skills & Culture | 2.2 |
| Governance & Accountability | 2.3 |
| Data & Tools | 2.1 |
| Vision & Strategy | 2.2 |
Source: Spark AI Maturity Score, Autumn/Winter 2026/27, n=108.
38% of agencies are still at Stage 1, while the most advanced self-reported score is at 3.8
Stage 2 is the biggest group, at 49%, and 13% have reached Stage 3. We haven't recorded any agency at Stage 4 yet – where AI changes what the agency sells and how it charges.
More than four in ten Stage 2 agencies are still at Stage 1 on at least one dimension. Progress is uneven, with agencies moving ahead in some areas and still starting out in others.
Take a harsher view, where one Stage 1 dimension puts the whole agency at Stage 1, then around six in ten would be at Stage 1.
Swipe to see the full chart →
The bigger the agency, the more uneven its AI maturity
Maturity in AI governance rises with size, from 2.2 in micro agencies to 3.0 in large ones. So does the gap between dimensions, from 0.1 to 0.7 points – and the weakest dimension holds the whole agency back on its AI maturity. Smaller agencies have the agility advantage to make progress on all four dimensions of AI maturity together.
Swipe to see the full chart →
Leaders scoring on their own rate their agency about 0.4 higher
This year we opened the Score on our website, and most of those scores come from one leader rating their own agency. In our client agencies, where the whole leadership team answered before an AI transformation programme, the average was about 0.4 lower.
That suggests self-assessment bias, so the real picture is likely to be a little lower than these figures.
One leader vs a whole leadership team
Average Spark AI Maturity Score, 1–4
See the data
| One leader, self-reporting on our website | 2.2 |
| Whole leadership team, averaged (client agencies) | 1.8 |
Source: Spark AI Maturity Score, Autumn/Winter 2026/27. Website scores come from single leaders. Client scores are the average of each leadership team, taken before a Spark programme.
Where does your agency sit?
Free, five minutes, with a tailored report on what to focus on next.
Chapter 01The productivity high
Everyone's faster. Few are better off
Two years ago, most agencies were only dabbling with AI. Now it's part of the working day. 90% of the agency staff we surveyed use AI at work, and 91% of AI users say it saves them time. That's a big win for individuals – but those time savings aren't yet consistently showing up on the agency's bottom line.
The share of staff saving time has doubled since 2025
45% of the AI users we surveyed saved time in 2025. Now it's 91%. And 90% of agency staff use AI at work.
The share saving time has doubled since 2025
% of AI users who save time each week
See the data
| 2025 | 45% |
| Spring 2026 | 89% |
| May–Oct 2026 | 91% |
Source: Spark AI pre-programme staff survey, three waves: Feb–Oct 2025, Oct 2025–Apr 2026 and May–Oct 2026. Base: respondents who use AI, n=219 (May–Oct 2026).
13% of AI users now save five or more hours a week
That's up from 5% in 2025, and 4% now save more than 10 hours. Only 9% save no time at all, down from 55%.
Models can now produce usable documents and presentations, so each use covers more of the task.
of AI users save five hours or more a week – up from 5% in 2025
save no time at all – down from 55% in 2025
Source: Spark AI pre-programme staff survey, May–October 2026. Base: respondents who use AI, n=219.
Why personal gains drain away
The time gets swallowed – back into emails, meetings and more of the same, faster. And the know-how stays with the person who worked it out.
Someone who has worked out a good workflow has a personal habit. An agency that has written it down, trained others on it and priced it into a proposal has a capability.
The appetite to learn AI is clearly there. The organisations seeing the greatest impact are building on that enthusiasm with structured learning, practical application and shared ways of working that help capability scale across the business.
Caroline MoffatL&D Business Partner, Publicis
Where does the value go?
The individual – for now
- YesThe individual
- Not yetThe client
- Not yetThe agency
- Not yetJunior talent
Map where AI fits in each role – and close the fluency gaps you find
- Audit your role with AI – a prompt to run today
- Check your AI fluency against the 4Ds
Spark AI Maturity Model® dimensionDimension: People, Skills & Culture
Chapter 02Where does the time go?
Saved time leaks unless you spend it on purpose
So people are saving time. We asked agency staff where that time goes.
Where saved time goes depends on team habits and expectations. The Spark AI Maturity Model® dimension of People, Skills & Culture is about building shared skills and giving people room to put them to use.
Where agencies sit – People, Skills & Culture maturity score
- Stage 1 · ExperimentationSelf-taught, uneven skills39%
- Stage 2 · AdoptionShared training, sharing becomes routine40%
- Stage 3 · OptimisationRole-specific development, time protected17%
- Stage 4 · TransformationAI-native, continuous by default4%
Source: Spark AI Maturity Score, Autumn/Winter 2026/27. Highlighted: where most agencies sit.
48% use AI to do more of the same, faster – only 9% use it for new kinds of work
Nearly half (48%) say AI helps them produce more in the same time. But greater productivity doesn't automatically mean greater profit, particularly when agencies deliver more for the same fee.
Saved time can also improve the quality of work (41%) or reduce pressure on teams (25%). Both are valuable. The challenge is making these gains count. With 45% saying saved time gets absorbed or is hard to account for, agencies need to decide where that extra capacity goes.
Only 9% are using saved time for new kinds of work
% of respondents – more than one answer allowed
See the data
| More output, same time | 48% |
| Absorbed or hard to account for | 45% |
| Better-quality work | 41% |
| Less workload pressure | 25% |
| New kinds of work | 9% |
Source: Spark AI pre-programme staff survey, May–October 2026, n=243. Respondents could select more than one outcome, so percentages don't total 100%.
37% of agencies still rely entirely on self-taught AI
To turn individual skills into agency-wide capability, agencies need to document what works, train the whole team and put someone in charge.
The conversation about AI tends to focus on efficiency and margin. I’d like us to also ask what it gives back to people. Agency life can mean long hours, and AI is a real chance to return time and headspace to our teams, but only if we choose to protect it rather than fill it.
Jo HiggsEvent Concept
72% haven't structured their knowledge for AI
Years of expertise, client insights and proven ways of working are buried in agency files and conversations. Making that knowledge available to AI could turn it into a shared resource that makes everyone’s work better.
How agency knowledge is held
% of agencies
See the data
| Informal, in people's heads | 32% |
| Agreed source for key documents | 40% |
| Structured and current for AI to use | 21% |
| Judgement encoded into AI systems | 6% |
Source: Spark AI Maturity Score, Autumn/Winter 2026/27, n=108. Single answer.
Swipe for more
still rely entirely on self-taught AI
have a named AI lead or taskforce driving adoption
have no named owner to drive AI adoption
Source: Spark AI Maturity Score, Autumn/Winter 2026/27, n=108.
Self-taught AI is how agencies end up running at two speeds. General training gives everyone a shared baseline, role-specific development helps people apply it to their own work, and continuous learning keeps that capability current.
See our AI training programmes
Where does the value go?
The individual first, and the client. The agency once AI is part of every workflow.
- YesThe individual
- YesThe client
- Not yetThe agency
- Not yetJunior talent
Run a team AI reset day – and turn your best methods into shared skills
- How to run a team AI reset day
- The one-hour weekly habit that separates agencies that move
Spark AI Maturity Model® dimensionDimension: Governance & Accountability
Chapter 03Who's keeping it safe?
A policy nobody checks is just a PDF
As AI becomes part of everyday work, agencies need to be clear about how it should be used. That means setting standards for client data, checking AI-generated work and being transparent with clients. Without clear policies and accountability, individuals are left to make those decisions for themselves.
Where agencies sit – Governance & Accountability maturity score
- Stage 1 · ExperimentationNo policy, down to the individual31%
- Stage 2 · AdoptionPolicy exists, adherence unchecked40%
- Stage 3 · OptimisationPolicy checked, formalised in client contracts26%
- Stage 4 · TransformationProactive governance, a commercial asset3%
Source: Spark AI Maturity Score, Autumn/Winter 2026/27. Highlighted: where most agencies sit.
66% of agencies don't actively check how AI is used
44% have an AI policy but don't check it's followed. A further 22% leave AI use unmonitored. Only 34% actively apply or review their rules.
For most agencies, governance has reached Stage 2 – Adoption. It isn't yet part of everyday practice.
How AI governance is applied
% of agencies
See the data
| How AI governance is applied – Unmonitored | 22% |
| How AI governance is applied – Policy, not checked | 44% |
| How AI governance is applied – Actively applied or reviewed | 34% |
Source: Spark AI Maturity Score, Autumn/Winter 2026/27, n=108.
What's at stake
- Client data Free consumer tools can use what you feed them to train their models.
- Ownership Only 25% of agencies have updated their contracts to reflect AI use in client work (The Wow Company, Benchpress 2026).
- Disclosure Since August 2026, EU transparency rules apply to certain AI-generated content, with marking requirements for existing tools following in December.
- Automation bias We tend to trust polished output, even when it's wrong, and AI trained on historic data repeats its stereotypes. Every piece of AI-assisted work needs a named person checking it.
- Your people In Spring, our Spark Report found that interest in IP and risk management had grown 50% in six months. Staff want guardrails. Without them, they carry the agency's legal and ethical exposure alone.
Governance can win work
Clients now ask about AI governance in pitches. 13% of agencies already count their AI capability among the reasons clients choose them.
Where does the value go?
Everyone – and the agency most of all
- YesThe individual
- YesThe client
- YesThe agency
- Not yetJunior talent
Put five simple rules in place this month – and check work before it goes out
- Minimum governance in five steps – no 40-page policy needed
- The EU AI Act check before work goes out
Spark AI Maturity Model® dimensionDimension: Data & Tools
Chapter 04Bolted on or built in?
Faster steps, same old process
Now to the work itself. Once agencies have shared ways of working and clear rules in place, they can start redesigning how work gets done. But most haven’t reached that point yet.
Redesign starts with what’s now possible
Adding AI to an existing workflow makes the same process faster. Redesign starts by asking what AI can now make possible. Which steps are no longer needed? What could be done differently? What new services could you offer clients?
That’s where the opportunity goes beyond saving time to creating new value.
Where agencies sit – Data & Tools maturity score
- Stage 1 · ExperimentationPersonal tools, no data structure33%
- Stage 2 · AdoptionCompany accounts, single source of truth48%
- Stage 3 · OptimisationConfigured assistants, structured knowledge17%
- Stage 4 · TransformationConnected systems, agency second brain2%
Source: Spark AI Maturity Score, Autumn/Winter 2026/27. Highlighted: where most agencies sit.
Only 9% have redesigned their workflows
Most agencies are still using AI within existing processes. 22% use it for productivity tasks, while 53% use it to improve client work. Another 16% have configured AI assistants, but only 9% have redesigned workflows end to end, with defined roles for people and AI.
Only 9% have redesigned the workflow end to end
How AI is used within workflows, % of agencies
See the data
| Productivity tasks | 22% |
| Improving client work in the existing process | 53% |
| Configured AI assistants | 16% |
| End-to-end redesigned workflows | 9% |
Source: Spark AI Maturity Score, Autumn/Winter 2026/27, n=108.
The commercial case for redesign
The potential gains are significant. One agency leader at Cannes reported a 1.2x productivity improvement from adding AI to existing workflows, compared with 6x after redesigning them.
McKinsey’s research points in the same direction. Nearly three-quarters of companies reporting AI-driven profit gains of 5% or more had redesigned their workflows, compared with around a quarter of other companies.
Source: McKinsey, The State of AI, 2026 (1,719 respondents, all industries). Not Spark AI data. The Cannes example is one agency leader’s experience.
Agents can help put redesigned workflows into practice
Agents have been one of the big developments since our Spring report. They can take on parts of a redesigned workflow, connecting systems, completing tasks and handing work back to people when needed.
The building blocks are simpler than they sound. An AI assistant has three. An agent adds three more, allowing it to act when triggered and work across systems.
Swipe to see the full chart →
An AI assistant Starts in a chat, answers in a chat
- 1InstructionsWhat you want it to do
- 2KnowledgeWhat it needs to read first
- 3StandardsWhat good looks like
+ Three more blocks = an AI agent Runs itself, works in your systems
- 4TriggersWhat starts it
- 5ConnectorsWhich systems it reaches
- 6OutputsWhere the result goes
Where does the value go?
Everyone, when the work changes
- YesThe individual
- YesThe client
- PartlyThe agency
- Not yetJunior talent
Rebuild one workflow from scratch – and hand the repetitive steps to an agent
- A blank-slate workflow redesign
- A checklist to build a competitor-watch agent
Spark AI Maturity Model® dimensionDimension: Vision & Strategy
Chapter 05Faster should cost more
Hourly pricing turns every efficiency into a discount
Now to pricing – how agencies charge for work that AI makes faster. You invest in AI, rebuild a workflow and free up time. Then the client says: if it took half as long, shouldn't I pay half as much? It isn't fair. But if you price by the hour, you built that trap yourself.
Where agencies sit – Vision & Strategy maturity score
- Stage 1 · ExperimentationNo AI strategy exists37%
- Stage 2 · AdoptionGoals agreed, not written down43%
- Stage 3 · OptimisationDocumented strategy, commercial model shifting19%
- Stage 4 · TransformationAI is a competitive differentiator1%
Source: Spark AI Maturity Score, Autumn/Winter 2026/27. Highlighted: where most agencies sit.
67% of agencies have no documented AI strategy
Most agencies recognise AI's potential, but only 33% have written down what they want it to achieve. Without that, it's hard to connect changes in the work to the value the agency wants to keep. The right answer varies too – a media agency's commercial model isn't a creative agency's.
Two-thirds of agencies don't have a documented AI strategy
% of agencies
See the data
| Is the AI strategy written down? – No documented AI strategy | 67% |
| Is the AI strategy written down? – Written strategy | 33% |
Source: Spark AI Maturity Score, Autumn/Winter 2026/27, n=108.
Only 6% have created new AI-enabled revenue models
AI is now inside the creative work itself: 50% of agency staff use it for ideas, creative development or visual work. Yet 31% of agencies still price purely on time. 35% price by output, 28% on value, and only 6% have a new AI-enabled revenue model.
If you price by the hour, every efficiency you find becomes a discount you hand over.
Your clients are getting your AI gains for free
How agencies price and sell their work, % of agencies
See the data
| Output or deliverable | 35% |
| Input or time-based | 31% |
| Value-based | 28% |
| New AI-enabled revenue model | 6% |
Source: Spark AI Maturity Score, Autumn/Winter 2026/27, n=108.
Where does the value land?
When AI creates value in your agency, what kind is it – and who ends up with it?
Time back
Hours saved on the same work. Margin and capacity – but only if someone captures it.
Ask: where did last month's saved hours go?
New revenue
Work you used to outsource, now done in-house. New services. A premium for speed to market.
Ask: what do we charge for now that didn't exist a year ago?
Given away
Faster becomes cheaper, because price is tied to time. In every other industry, faster costs more.
Ask: did the client pay less, or did we do more for the same?
Client value
Speed to market, better decisions, a client who looks good to their board. Makes you harder to replace.
Ask: can the client see it – and can they name it?
Source: Spark AI.
55% of agencies have no formal client position on AI
Pricing changes are hard to sustain unless clients understand how AI is used and what's better for them as a result. Their boards are asking what AI is saving them, and marketing teams pass that question to their agency. Give them evidence they can take upstairs and you become harder to replace.
How agencies discuss AI with clients
% of agencies
See the data
| How is AI discussed with clients? – Not acknowledged | 14% |
| How is AI discussed with clients? – Discussed informally | 41% |
| How is AI discussed with clients? – In contracts or terms | 32% |
| How is AI discussed with clients? – Part of why clients choose them | 13% |
Source: Spark AI Maturity Score, Autumn/Winter 2026/27, n=108.
Where does the value go?
The agency – if it prices for value
- YesThe individual
- YesThe client
- PartlyThe agency
- Not yetJunior talent
Price your next project on value – and show clients the thinking behind it
- Three steps to price one new project differently
- The pricing ladder, from output to outcome
Spark AI Maturity Model® dimensionDimension: People, Skills & Culture
Chapter 06Junior roles evolving
Redesign how they learn
Now to junior talent – what AI means for how agencies hire and grow juniors. Redesigning workflows can open up new commercial opportunities. Our hypothesis is that it can also give junior talent a stronger role.
Where agencies sit – People, Skills & Culture maturity score
- Stage 1 · ExperimentationSelf-taught, uneven skills39%
- Stage 2 · AdoptionShared training, sharing becomes routine40%
- Stage 3 · OptimisationRole-specific development, time protected17%
- Stage 4 · TransformationAI-native, continuous by default4%
Source: Spark AI Maturity Score, Autumn/Winter 2026/27. Highlighted: where most agencies sit.
Under-25s in IPA member agencies fell 19.2% in 2025, according to the IPA Agency Census
employees aged 25 and under across IPA agencies
vacancies across all levels
agencies employing graduates or apprentices
AI is one factor among wider commercial pressures, but junior roles are especially exposed, because AI can now do many of their traditional tasks.
Source: IPA Agency Census 2025.
Cutting junior hiring saves money today. It also leaves fewer people with the experience agencies will need later.
Juniors usually learn by doing the early, lower-stakes parts of a job. AI now does many of those tasks, so this way of learning has to change.
Our hypothesis: the first time a junior takes on a new workflow, a senior works through it with them and they make the decisions together. Those decisions get captured in the AI, so the junior needs less support on the second project, and less again on the third.
The most exciting conversations are about things clients have never been able to do before. That's where AI opens up new budgets, rather than squeezing the ones we already have. It's also where the next generation of talent will learn most – working on what's new, not just what's routine. Whatever an organisation's AI strategy is, they must recognise the value young talent brings to any business and invest time and money into developing the skills that will power their growth now and into the future.
Graeme GriffithsDirector of Media Research, IPA
Where does the value go?
Junior people and the agency
- YesThe individual
- YesThe client
- YesThe agency
- Could beJunior talent
Give a junior a whole piece of work, with a senior alongside – and capture what they learn as AI Skills
- A plan for a junior's ownership of one output
- How to turn the decisions into reusable AI Skills
Conclusion
The agency of the future
So, where's the value going? Right now, to individuals and clients. For most agencies, those gains aren't yet turning into a commercial return.
All of that can and will change. Last year most agencies were at Stage 1. Now Stage 2 is the biggest group. The time AI frees up is most valuable when it goes back into client relationships and judgement. That's what clients buy when they buy your thinking.
Swipe to see the full chart →
- Chapter 1People save timeAnd use it to add more value for clients and do new things
- Chapter 2The team shares what worksDrains away if know-how stays with one person
- Chapter 3Rules make it safe to scaleDrains away if nobody checks
- Chapter 4Workflows are redesignedDrains away if AI is only bolted on
- Chapter 5Pricing captures the valueDrains away if you still bill by the hour
- Chapter 6Juniors learn on the new workDrains away if you stop hiring them
Each one makes the others more valuable
Where to start
How Spark AI can support your AI maturity
Spark AI is the AI performance partner that builds organisation-wide AI capability for agencies and brand teams.
AI transformation, training and tools, all built on the Spark AI Maturity Model®.
We've worked with more than 90 agencies and brand teams across the UK and US, from independents to global networks. We also teach on the AI for Business programme at Oxford's Saïd Business School, and wrote the #1 bestselling book Shift – AI for Agencies.
Spark AI Accelerator clients
A full stage up
in ten weeks on the Spark AI Maturity Model® – the front of the pack with AI today.
The AI Accelerator is where transformation starts. Spark Frontier then works alongside you as you move up the model.
What changed for three agencies
Before and after our AI Accelerator.
Digital brand and experience agency
25 people, London
Maturity score1.5 → 2.5
- Faster work
- Proposals in 3 days, not a week. RFPs in half the time.
- Capability
- Staff rating themselves highly capable up from 24% to 75%.
- Money
- AI product shots sold on value, not hours.
- Rules
- AI skills part of appraisals, plus a weekly AI hour.
Digital product agency, fitness and wellness
20 people, London
Maturity score2.5
- Faster work
- Grant application time cut to a third.
- Capability
- 86% use AI daily. 1 in 5 build tools colleagues use.
- Money
- A new AI product pitched on outcome-based pricing.
- Rules
- Shared prompt library with review and version control.
Global branding studio
300 people, London and New York
Maturity score2.6
- Faster work
- Licensing checks in 3 steps, not 7.
- Capability
- Every creative role uses AI tools daily.
- Money
- Visualisation brought in-house.
- Rules
- AI terms added to 14 client contracts, with a named AI lead.
1 Step one
1Start with the free AI Maturity Score™
It diagnoses where your agency sits on each dimension – and which programme to start with. 12 questions, five minutes, no sales call.
Get your AI Maturity Score™- Your stage on each of the four dimensions
- How you compare with the agencies in this report
- A tailored report on what to focus on next – and which programme fits
2 Step two
2Pick the route that sounds most like you
If you're…
Starting out, with no shared approach yet
AI Foundations
A half-day workshop that gets the whole team to a shared baseline.
Explore AI FoundationsIf you're…
Experimenting without a purpose
AI Accelerator
Ten weeks of role-specific learning built around real client work.
Explore the AI AcceleratorIf you've…
Got the tools, but can't see it in the numbers
Agent Build Sprint, AI Tool Building and Commercial Model Sprint
Redesign a workflow in two days, build AI tools around how you work, and move your pricing towards value in four weeks.
Explore the Agent Build SprintExplore AI Tool BuildingExplore the Commercial Model SprintIf you're…
Worried about momentum to innovate
Spark Frontier
Monthly support that keeps your AI programme moving as the technology changes.
Explore Spark FrontierNot sure? Email hello@wearespark.ai and we'll talk it through.
Get the full report – and find out what to do about the findings
This page shows what we found. The full report shows what to do next.
On the live page, your title and form from the section below appear here.
Free download - The Spark Report A/W 2026/27
The Spark Report:
Where's the value?
Fill this in and it's yours straight away.
By sharing your email address you agree to receive helpful AI news from us. You can unsubscribe at any time.
About the research
How was the Spark Report research done?
Over 350 responses across two sources. The Spark AI pre-programme staff survey shows how agency staff use AI: 243 staff, May–October 2026. The Spark AI Maturity Score shows how agencies are organised around AI: 108 agencies and brand teams, August–September 2026. Maturity scores are reported at agency level: where several people from one agency took part, we used the average of their scores so no single agency is overrepresented. To check for self-assessment bias, we compared website scores with leadership team averages from six client agencies, collected before their programmes started. Some questions were added during collection, so each chart shows its own base. The research also draws on more than two years of work with more than 90 agencies and brand teams, and feedback from 2,500 agency professionals.
What is the Spark Report?
The Spark Report is Spark AI's twice-yearly research into how agencies and brand teams are adopting AI, and what is driving real value.
How many agency staff use AI?
90% of agency staff use AI at work, and 91% of those who use it save time – up from 45% in 2025 (Spark AI pre-programme staff survey, May–October 2026).
Why aren't agencies seeing a return on AI?
Most have sped up the old model instead of changing it. 91% of AI users save time, but only 9% of agencies have redesigned a workflow end to end and 6% have a new AI-enabled revenue model. Redesign the work, don’t just speed it up.
What is the Spark AI Maturity Model®?
The Spark AI Maturity Model® is a four-stage model – Experimentation, Adoption, Optimisation, Transformation – scored across Vision & Strategy, People, Skills & Culture, Data & Tools, and Governance & Accountability. The average agency scores 2.2 out of 4, early Stage 2. Stage 4 is as far ahead as we can see today. As AI develops, we expect to add further stages.
Do companies that redesign workflows get more value from AI?
Yes. Across industries, nearly three-quarters of companies where AI adds 5% or more to profit have fundamentally redesigned their workflows, against a quarter of everyone else (McKinsey, The State of AI, 2026). In agencies, only 9% have redesigned a workflow end to end (The Spark Report, Spark AI, Autumn/Winter 2026/27).
Is AI reducing junior hiring in agencies?
Junior roles are under pressure. Under-25s in IPA member agencies fell 19.2% in 2025, vacancies fell 40.8%, and the share of agencies employing graduates or apprentices dropped from 56% to 43% (IPA Agency Census 2025). In the US, employment among 22- to 25-year-olds in AI-exposed occupations is 19% below where it would be if it had kept pace with less-exposed roles (Stanford Digital Economy Lab). AI is one factor among wider commercial pressures.
Are clients moving away from hourly agency fees?
Yes. Among 69 multinational advertisers, labour-based agency fees fell from 54% in 2011 to 17% today, while fixed-fee and output-based models grew from 20% to 35% (WFA and Agency Mania Solutions, via The Media Leader). Yet 31% of agencies still price purely on time, and only 6% have a new AI-enabled revenue model (The Spark Report, Spark AI, Autumn/Winter 2026/27).
Is the full report free?
Yes. Fill in the short form on this page and get the full report straight away, including a practical plan for every chapter.
Written by
Read our bios on the team page →
Thanks to friends of Spark for supporting our data collection:
