We’re in the 2008s of AI
I called social media when everyone said it was a fad. I am watching the exact same pattern play out again, only this time it reaches deeper, into how we work and how our customers experience us, not
Let me take you back.
When I started out in social media marketing, normal people used social media. They did not market on it. The idea that you would build a business function, let alone a career, around Facebook and Twitter struck most serious people as faintly ridiculous.
The consensus was clear. Social media marketing was a fad. These little apps would never compete with television. At best they would be a small bolt-on to a real marketing department, a junior posting things nobody read.
I did not see it that way, and I was right.
What almost nobody clocked at the time was where attention had actually gone. People were paying far more attention to these channels than to TV. Even when they were watching television, they were watching social media at the same time, phone in hand, eyes down. The channels themselves were going to explode. And an entire industry was being built around a thing everyone had filed under bolt-on.
That bolt-on became arguably the biggest marketing channel on earth.
And then the ripples started, and they did not stop.
New jobs appeared. The social media manager did not exist, and then suddenly every company needed one. A whole job title, invented inside a decade.
A whole new industry exploded out of it. The creator economy. People building their entire income, and their entire lives, around social platforms. Not marketing for someone else. Marketing as the business itself.
The money followed the attention. Companies now spend more on social than on almost any other channel. It is the lowest-hanging fruit there is for a small business, the first place you go, not the last.
Nobody realised how enormous this was going to get. It rippled through every business I have ever talked to, and it changed all of our lives independently of work too. The thing everyone called a fad rewired marketing, created industries, and reshaped how a generation spends its day.
I was 21, running social for ODEON Cinemas, doing a job that had barely been named yet. I was early. Being early was the whole point.
I have seen this pattern before. I am seeing it again.
The parallels between social media then and AI now are enormous, and once you have lived through one of these, the second one is impossible to un-see.
The platforms. Back then it was Facebook, Meta, Instagram, Twitter, Snapchat. The seismic shift came from a handful of platforms that everyone used. Today it is Claude, OpenAI, Gemini, and the same thing is happening. A small number of platforms, adopted at a pace that makes the last cycle look slow.
Personal and professional at once. Social media worked because it was both. You used it at home and you used it at work, and the two fed each other. AI is exactly the same. People are writing their nutrition plans, their fitness and transformation plans, their personal admin with it on a Sunday, and running their actual job with it on a Monday. It is a utility now, used every single day, just as social media became. And just as social quietly became the place customers expected to find you, AI is quietly becoming the way customers expect to interact with you.
The new industries. Last time it was social media managers and the creator economy. This time it is AI consultants, AI trainers, and independents building businesses on top of these tools. The job titles are being invented in real time, again.
The stickiness. Be honest about this one. These tools have dopamine built in, the same as social did, the little hit that keeps you coming back. That is not an accident, and it is part of why adoption moves the way it does.
The early adopters are already winning. Just as they did in the 2010s. The level of change is incredibly early, and the gain on offer is enormous, and most people have not noticed yet.
Here is the part I would put real money on. Just as a 21-year-old social media manager turned out to be the start of an entire profession, there will be AI managers, AI consultants, and AI leads inside organisations who become the powerhouse of knowledge over the next five to ten years, without anyone quite realising it happened. The person quietly 10x-ing their output today is the person running the function tomorrow.
What is different this time
This is the first time I have seen the same pattern point at a completely different use case, and the difference matters.
Social media was, at its core, marketing. It was about how you communicate outward and win customers, one channel, one edge of the business. AI is broader than that. It changes how the work itself gets done on the inside, the production, the productivity, the output. And it changes the product your customers actually touch on the outside, the interface, the service, the personalisation. Two fronts, not one.
So the surface looks identical, the big platforms, the explosion of new skills, the seriousness organisations are starting to attach to it, some doubling down and investing hard while others sit back sceptical. But underneath, the change is deeper. Social media changed how you reach people. AI changes how you operate and what you offer them when they arrive. That is a far more seismic shift, because it touches the core of the company and the edge of it at the same time.
And the timing rhymes too. The pace of change always takes longer than people think. Social media built through 2008, 2009, 2010, and only truly exploded in the 2020s. AI will follow a similar arc, just faster, because we have all been through this cycle once and we recognise the shape of it now.
So where are we? We are in the 2008s. Maybe just past it. The stage of niche blogs and business weirdos championing the thing while everyone else waits to be convinced. I was one of those weirdos last time. I am happy to be one again.
The next ten years will bring more change, at more scale, than the last cycle did. This piece is how you actually adopt it into your company, written by someone who has seen this pattern twice and has always believed in pulling new technology into an organisation as early as physically possible, because early is where the gain lives.
Built in, not bolted on
Here is the first lesson the social media cycle teaches you, applied to AI.
There are two ways to bring this technology into a company. You can build it in. Or you can bolt it on. The companies that treated social as a bolt-on lost a decade to the ones that built it into how they operate. The same fork is in front of you now.
Start a company today and you start with nothing to undo. No legacy systems. No legacy customers. No legacy data sitting in seven places. No process someone built in 2015 and now defends with their career. No sacred way of working everyone tiptoes around.
This is the single biggest advantage in business right now.
Look at what an AI-native team does from day one:
They build the workflows in, not on. You could never bolt deep AI automation into the spine of a company that already has tons of legacy, tons of customers, tons of data. The cost of rewiring is too high and the risk of breaking what works is too real. A native team wires it in on day zero, before there is anything to break.
They build it into the product, not just the back office. A native team does not only automate its own workflows. It ships a conversational interface, real personalisation, and AI-native customer service from day one, because there is no legacy product to retrofit. The customer feels the difference immediately, not in version four.
They collaborate around shared, automated skills. Not just prompts in a doc. Actual reusable skills that one person builds and the whole team switches on. The knowledge of your best operator becomes something everyone else can adopt.
One AI-enabled person does the work of ten. This is the line everyone quotes and almost nobody builds for. A small native team is not a smaller version of a big company. It is a different shape entirely. Fewer people, more output, more margin, more speed.
The mindset is already won. Everyone in the building chose to be there. They are advocates by default. You are not spending eighteen months dragging an organisation across a line it does not want to cross.
That last point is the quiet one, and it might be the biggest. We will come back to it.
Quality of output is a craft, not a magic trick
Here is the thing native teams understand that retrofitters keep getting wrong.
The first one, two, even three iterations of most AI output are not good enough to ship.
That is not a flaw. That is the job. The skill is not typing a prompt and pasting the answer. The skill is knowing the first answer is a draft, building the loop that takes it from draft to shippable, and knowing when to stop.
Native teams design for that loop. They treat AI like a junior who is fast, tireless, occasionally brilliant, and needs direction. Retrofitters treat it like a vending machine, get a mediocre result, and conclude the whole thing is overhyped. Same technology. Different craft. Same mistake the TV-only marketers made when they posted once, got no likes, and decided social did not work.
And the stakes jump the moment that output faces a customer. A rough first draft is fine inside the building. The same rough draft sent to a customer, or spoken by the assistant on your website, is a brand event. So the teams that win build the iteration loop in hardest exactly where the customer can see it, the support reply, the product recommendation, the on-site answer, because that is where a weak output costs you trust, not just time.
AI has to do a job, or it is theatre
Let me be blunt about why any of this matters, because there is a lot of AI activity right now that is just activity.
The use case for AI is not the tool. It is the business objective the tool serves. There are only three that count:
Drive faster growth at the top line.
Drive better profitability.
Increase operational efficiency to lower cost.
That is it. Everything else is a hobby with a budget.
The whole thesis sits on a handful of outcomes:
driving faster growth,
driving better internal productivity,
5x-ing the output of your people from day zero,
building those people in so profitability grows with them,
and giving customers a faster, more personal experience than your competitors can match.
If your AI programme cannot be traced back to growth, profit, or efficiency, you are doing AI theatre. Stop, and start again from the objective.
The second front: your customer experience
Everything so far has been about the inside of the company, about how your people work. That is half the story. There is a second front, and most internal-focused AI programmes walk straight past it.
AI is not only changing how you operate. It is changing what your customers experience, and it already is, quite frankly, in the products you use every day. This is not about individuals firing up an LLM at home. This is about companies building AI products and features into their own offering, and it is another shift you have to get your head around and double down on. I see it everywhere, in the products I use, and in one of the companies I run today, JAAQ.
If the internal shift is the part of AI with no real precedent, this is the part that rhymes most directly with the social media cycle, except it goes further. Social media changed how you reached your customer. This changes the experience your customer actually has with your product.
Here is where it is showing up.
The interface has gone conversational. Before AI, almost every business interface was a workflow. Click, read, filter, click again, find the thing, buy the thing. You still have to do most of that today. Layer AI on top, and the customer can simply ask. Picture an e-commerce business where, instead of fighting filter menus, you say:
find me a black T-shirt under £10,
find me the best-reviewed mower for my garden,
find me the most-used foundation in the range.
You talk to the company to make the decision. We should be enabling this far further than we are. The shop becomes a conversation, not a maze, and the friction that quietly kills conversion starts to disappear.
Conversational AI is arriving fast in health. This is the front I know best, because we are building it. At JAAQ we are building conversational AI, clinically governed AI and data, so people can better engage with their health, the health of their mind and the way they look at their physical health. And as a customer myself, I live the other side of this. I use WHOOP every day, and I can now talk to it. I can ask it questions grounded in exactly how it measures my body’s reactions, and get far deeper analysis of my sleep, my training, my stress, the tweaks I should be making to my schedule. Having a real conversation with something that is monitoring your body is genuinely powerful, and it was science fiction a few years ago.
B2B is not exempt. Visit a business tool today and you can increasingly ask it questions directly, raise a service issue, and have that issue handled conversationally from start to finish. A whole layer of support, onboarding, and sales enablement is being rebuilt around conversation instead of forms and ticket queues.
Recommendation engines are being reborn. Personalisation engines used to be brutally hard to build. Genuinely complex, expensive, specialist work that only the biggest companies could afford to do well. AI changes the economics of that overnight. The model can learn far more about a user from the data pool you already hold, and recommend with real intelligence. The next product if you are Amazon. The next thing to watch if you are Netflix. The next thing a customer did not even know to look for. The depth of personalisation on offer now is something we have simply never had before, and it is only going one way.
Pull it together and the point is simple. Using these tools to automate internal tasks is great. Using them to automate and improve the customer experience at scale is another enormous lever, on both revenue and efficiency, in exactly the business cycle we are in right now. The companies that win will run both fronts at once, the inside and the outside, and they will not treat the customer-facing one as a feature they get to later.
Most of us are not native
Here is the uncomfortable part. Almost none of us reading this run an AI-native company. We run real businesses with real legacy, and we cannot start again. Same as the established brands in 2010 who could not suddenly become digital-first. So the question is not whether native teams have an edge. They do. The question is what you do about it from where you actually stand.
Take two companies.
The 100-person company. $50m in revenue. Already profitable. Growing nicely. You might argue there is no problem here. But every market on earth is being disrupted, and the macro backdrop is brutal: rates higher than most operators have ever traded through, cost of living biting, inflation still in the system. If you want to grow faster than your market would naturally hand you, you have to adopt this technology, inside and out. The competitor who lets customers simply ask for what they want, and answers in seconds, will quietly take your business while you are still proud of your filter menus. Comfortable is not safe.
The 10,000-person company. Highly profitable. Highly predictable. Publicly traded. Almost no obvious problems, and a board that wants profit to keep climbing anyway. Different scale, same goal. Neither company wants to blow up the status quo. Both are fighting the same war for talent, paying up for the best people. Both are sitting on enormous legacy processes, run by humans, that are wildly inefficient and completely untouched by AI. And both have legacy customer experiences, clunky portals, call-centre queues, one-size-fits-all messaging, alongside legacy sales and marketing motions, all of which will get eaten by whoever rebuilds them around AI first.
This is the most competitive market I have operated in, and the hardest for talent I have ever seen. New businesses are appearing constantly. The bar is simple and unforgiving: are you growing the top line faster, or are you making more profit? AI is now one of the few real levers on both.
The real fight is mindset
Before any tool, any audit, any rollout, you have to deal with people. Because the biggest blocker to AI adoption, at 100 people or 10,000, is not technical. It is human. It always is. The hardest part of the social media shift was never the software either. It was getting people to believe it mattered.
Add the science. People do not like change. That is not a character flaw, it is wiring. So the move is not to fight the fear of change. The move is to enable people through it, with training and a proper level-set in the mind before you ask anyone to do anything differently.
And the fear here is sharper than normal change fear, because it is attached to identity. People tie AI to their own skills, their talent, their sense of being good at the thing they do. Really good people feel it. Average people feel it most of all, and that is the honest bit nobody says out loud. Everyone, deep down, is protecting their job.
There is a second fear too, and this one points outward. Leaders freeze at the thought of AI speaking to a customer in their brand’s voice. What if it gets something wrong, in public, with their logo on it. That caution is healthy. Used as a permanent excuse, it is fatal, because your competitors are getting comfortable with it right now. The answer is not to keep AI away from customers. It is to govern it properly, keep a human in the loop where it matters, and let it earn trust in the low-risk places first.
So let me be clear about where I stand.
I do not advocate using AI primarily to make people redundant.
What I do believe is powerful is this. AI creates a level playing field for how people adapt to a massive change in how work gets done. You cannot get away from the change. The people who adopt it are the ones who win their careers over the next ten years. The ones who fight it, or genuinely cannot adapt to it, are the ones who lose. That is not a threat. That is just the shape of the decade, and we watched the exact same sorting happen to everyone who refused to take social seriously.
And while leadership is busy deciding whether to engage, the organisation has already started without them. That is shadow AI, people running unapproved tools because the early adopters worked out months ago that they can 10x their output and make their work life better. It is a danger on security and on data. It is also the loudest possible signal that the demand is already there. Your job is to bring it into the light, not pretend it is not happening.
The playbook
So how do you actually do this, in a real company, with real people, without it turning into a deck nobody reads?
Here is the sequence I would run.
Step one: audit the capability before you touch anything
You cannot fix what you cannot see, and most organisations have no honest picture of their actual AI capability. They have a vibe. The vibe is wrong, usually optimistic at the top and chaotic underneath.
A real audit needs two layers, and this is the core of my conviction: you need a product overlay and a people overlay. Technology alone cannot do it. Humans alone cannot do it. You need both, which is exactly the gap we are building into AcademyAI.
The product layer gives you the truth that people will not self-report:
Measure actual tool usage across the organisation. Who is using what, how often, how deeply. Not what they say in a survey.
Run a real capability diagnostic by role. Test what people can actually do, not what they think they can do. Self-reported skill is almost always wrong in both directions.
Map the current workflows by function. What are the tasks, who does them, how long do they take today. This becomes your baseline for everything you measure later.
Map the customer-facing touch points too, not just the internal ones. Where do customers search, wait, get stuck, contact support, abandon the basket. Those are the moments AI can change the experience, and they belong on the same map as your back-office workflows.
Surface the shadow AI. Find the unapproved tools already in use. These are gold, because they show you where the real demand and the real early adopters already are.
The people layer gives you the context the data cannot:
Survey the organisation, then go further and ask people informally, where they will actually tell you the truth.
Run internal workshops by function to surface the real friction and the real wins on the ground.
Interview managers on where their teams’ time actually goes, versus where it should go.
Map your early adopters, your fence-sitters, and your genuine resisters. You will lead each group completely differently.
The output is a capability map, by function and by seniority, sitting alongside a customer-journey map of where the experience breaks down today. Where you are now, against where the business needs you to be, on both fronts. That map is the foundation. Skip it, and everything downstream is guesswork.
Step two: base-level the entire organisation on fundamentals
Once you can see the picture, you put everyone through fundamentals. Everyone. The graduate and the CTO. No exceptions for seniority, because seniority and AI fluency have almost no correlation right now.
Fundamentals means people genuinely understand:
what these tools actually are and, roughly, how they were built,
what they are brilliant at,
what they are bad at and where they will confidently lie to you,
when you should reach for them, and just as importantly, when you should not.
That last pair matters more than people think. Knowing when not to use AI is a core skill, not a footnote.
Do this at scale, gamify it so people enjoy it and come out with something to show for it, a level, a badge, a real sense of progress. Make it a thing people want to do, not a compliance module they click through. Done with the right tools, base-levelling a whole organisation is genuinely achievable.
Be honest about the clock.
This phase can take weeks, months, or the better part of a year depending on size. That is fine. You cannot build the upstairs before the foundation is set.
Step three: capability labs, function by function
Now you go deep. For each function, you take the workflows you mapped in the audit, and you teach plus implement specific, practical automations against them. Not theory. Live work.
Examples of what that looks like in practice:
Data analysis. Cleaning messy data, building pivots, pulling insight out of a spreadsheet that used to take an analyst a day.
Sales outbound. Research, personalisation, and sequencing at a volume one person could never hit by hand.
Reporting. Automated weekly and monthly reporting generated straight off the raw data, instead of someone rebuilding the same deck every Friday.
End-to-end task automation. Stitching multi-step processes together so a whole workflow runs with one person supervising instead of three people executing.
Marketing. Generating campaign variations, ad units, and creative at a scale and speed that changes what testing even means.
Customer support. Drafting responses, triaging tickets, and turning a messy knowledge base into instant answers.
Customer experience. Building the customer-facing AI itself: the conversational interface, the recommendation engine, the personalisation layer people actually feel. This is the function that ships AI to the outside world rather than only using it on the inside, and it is where a lot of the growth hides.
Finance. Reconciliation, first-pass forecasting models, variance analysis explained in plain English.
Product. Synthesising user research, drafting specs, turning a rough idea into a structured PRD in minutes, and designing the AI features that go into the product itself.
Recruiting and HR. Writing job descriptions, screening at the top of funnel, prepping structured interviews.
Legal and commercial. First-pass contract review, redlining against your standard positions, drafting routine clauses.
Engineering. Code review, test generation, and documentation that nobody wanted to write by hand.
Meetings. Notes, decisions, and action items captured automatically, so the work of the meeting survives the meeting.
The point of a capability lab is not to admire the demo. It is to leave the room with a workflow that is actually faster than it was when you walked in.
Step four: advocacy, champions, and shared skills
This is where it scales or stalls, and the engine is peer-to-peer learning.
Every organisation already has AI champions. The early adopters from your shadow AI problem are your asset here. Harness them, two ways:
Get them sharing, openly and collaboratively. What they are building, what works, what failed. No hoarding, no quiet edge-keeping. Make generosity the high-status behaviour.
Get them building reusable skills. This is the multiplier. When a champion turns their best workflow into a packaged skill, anyone else can just adopt it. One person’s hard-won capability becomes a switch the whole team can flip. That is how knowledge share stops being a workshop and becomes infrastructure.
Run it through internal share-outs and workshops, and then do the thing most programmes forget: embed it in the culture, so it keeps running after the launch energy fades.
Step five: the executives have to walk it, not just fund it
A block that gets skipped constantly, and the skip is fatal.
The C-suite has to know how to use AI. Not the headline version. The real ins and outs, the genuine power, and the honest limits. They have to talk the talk and walk the walk, in their own work, in front of people.
Because a transformation led by executives who do not use the thing themselves is a transformation nobody believes. Staff can smell it instantly. If the leadership team is exempt from the change they are demanding, the change does not happen. Top-down credibility is not a nice-to-have, it is the permission structure for everyone below.
And it is not only the internal rollout they need to grasp. The biggest, riskiest bets, the decision to put AI in front of customers in the product itself, will land on their desk. You cannot weigh a risk you have never felt, or spot an opportunity in a tool you have never used.
The comms, inside and out
Through every step, the internal communication has to run two directions at once. Top-down, so there is air cover, clarity, and obvious leadership intent. And bottom-up, so you are harnessing the change already bubbling up from the early adopters rather than fighting it. Both directions, at 100 people or 10,000. One without the other dies.
And do not forget the comms pointed at your customers.
The moment you put AI into the product, tell them. Be clear about what the assistant can do, where a human is still in the loop, and how their data is handled. Set the expectation before the first conversation, not after the first complaint. Done well, this becomes a trust-builder and a differentiator, the same way the brands that explained social early built the biggest followings. Done silently, it becomes the thing that gets screenshotted. External communication about your AI is part of the product now, not an afterthought for the legal team.
How you know it is working
A transformation you cannot measure is a transformation you cannot defend at the next board meeting. So you need a framework, and it has to cover both fronts, the internal one and the customer-facing one. Four things to watch.
Productivity per person. The hard one to measure honestly, so use proxies. Track tool usage across the organisation, and track the reduction in time to complete specific tasks against the baseline you captured in the audit. You will not get a perfect number. You will get a clear direction, and direction is enough.
Revenue from internal output. The growth your own people create when they can suddenly produce more. This is where the internal business case lives:
Outbound volume that was simply impossible before. Send twice the outbound, surface many times the leads, create opportunities that did not exist. More opportunities tends to mean more sales, naturally.
Marketing reach and variety. More campaigns, more creative variations, more ad units, more tests running at once.
Shorter cycles. Faster testing, faster reporting, faster learning. Speed is its own form of growth.
Customer experience. The second front, and the reason it is so valuable is that it lands on both sides of the ledger at once. The revenue above comes from your people doing more. This is revenue and efficiency that come from the customer’s side of the glass.
On the growth side:
Conversion. A conversational interface that helps people find the right thing faster lifts the rate at which visitors become buyers. Friction is where conversion goes to die, and conversation removes friction.
Basket size and cross-sell. Genuinely intelligent recommendation lifts average order value, because the model surfaces the thing the customer did not know to look for.
Retention and churn. A more personalised, more responsive experience keeps people. For an engagement business like JAAQ, the depth and frequency of engagement is the product, not a vanity metric, and it is the leading indicator of almost everything downstream.
On the efficiency side:
Support deflection and cost to serve. Conversational AI that handles service inquiries end to end means fewer tickets reaching a human, faster resolution, and a lower cost to serve every customer you have.
Speed to answer and time to value. Customers get what they need sooner, which shows up first in satisfaction scores and eventually in loyalty.
The trap on this front is measuring activity instead of outcome. Do not report that the assistant had ten thousand conversations. Report that conversion rose, churn fell, and cost to serve dropped. Same discipline as everywhere else, tie it to growth, profit, or efficiency, or do not bother.
Cost savings. The honest part, told straight.
I do not advocate for AI as a cost-cutting play first. But pretending cost does not move would be dishonest. The reality is that some people will not adopt it. Some genuinely cannot manage the change, and they will become your underperformers. Meanwhile the people who are brilliant at it will create efficiency that is hard to ignore. Over time, organisations will very likely save on cost, in headcount and in the sprawl of tools they no longer need, and that is the profitability driver whether anyone likes saying it or not.
Two things can be true. You can refuse to run an AI programme primarily as a redundancy exercise, and you can be clear-eyed that efficiency will surface who is adding value and who is not.
So where does that leave you
I have watched this movie once already.
A technology everyone called a fad turned out to be the biggest shift in its field, created whole industries, invented new jobs, and quietly made the early adopters into the people who ran the place a decade later.
The only people who lost were the ones who waited until it was obvious.
We are in the 2008s of AI. The platforms are here. The skills are forming. The new job titles are being written as we speak. The difference this time is that AI does not just change how you reach people, it changes how the work gets done and what your customers actually experience, which makes it much bigger.
If you are building native, your advantage is real, but it is not permanent. Asymmetries get priced in. Build the workflows, the shared skills, the customer experience, and the culture in now, while the gap is wide.
If you are retrofitting, you are not too late, but you are on the clock.
The native teams are not waiting for your audit to finish. Your edge is not a blank sheet, you do not have one.
Your edge is a real business, real revenue, real data, and real people, pointed at the change with intent instead of dragged toward it.
I was 21 and early once, and being early was the whole point.
It is early again.
The question is not whether this becomes enormous.
It already is.