What happens to social when AI eats service?

I once did some consulting work for a big mobile phone company. They wanted to get their customers to love them. "Love" was a big word for the brand team. They were committed to love.

In a quieter moment I remember being frank with one of the client team. The company was unlikely to achieve love with its customers if, each year, those customers had to phone up and threaten to leave the relationship just to get a better deal. Annual threats of divorce aren't great grounds for a loving relationship. This brand ranked, and still ranks, low on customer satisfaction.

This week's Financial Times Weekend Magazine contains a really good article on this: How did customer service get so bad?

I've spent years thinking about this. I should confess - I'm a proper service nerd. If I have a book in me, it's one about service (and maybe it should stay inside me. After all, as neurodivergent special interests go, this one is very niche).

The article does a good job of summarising the basic causes of customer service failure. Lack of investment in customer service, lack of customer service skills, fragmentation of the service chain across myriad third party providers, all stacked against rising customer expectations.

But having spent a 25 year career building, studying and fixing services, and trying to work out why "the act of service" seems to be so incompatible with "services", I think the article misses a big point. That fundamentally, we don't value service.

Since 2000, I saw software eat service. The article is really a study of what that era has left us with. What interests me, now that I'm running an artificial intelligence business, is what happens when AI eats service.

Given our AI business is focused on social media, I'm particular interested in what happens when AI eats service on social media. Given social teams expect a 318% increase in the use of AI for customer support in 2024, it's an increasingly urgent question.

But first, before we can address how AI will impact service on social - for better or worse - we need to consider why we're already getting bad service.

We don't value service (enough)

Go to an established, locally-owned restaurant in Milan, and odds are the waiter will be a man in his late 50s. He'll appear effortless in his ability to read the entire room - spot the dropped fork at table 3, anticipate the bill request from table 7, sense that table 2 have been waiting too long - and manage them to a good conclusion.

Go to an equivalent UK restaurant and odds are your waiter is much younger and usually much less capable. This isn't a criticism, it's just a recognition that we see service sector jobs as low-grade - something to do whilst we look for a 'real job'.

The irony of course is that we tend to look down on people who work service jobs, whilst simultaneously relying on the service sector for almost every aspect of our lives.

We get bad service because, in advanced Western capitalist economies, we don't value service. And you really have to value service, because although in principle service is a simple and human dynamic, making it a reality in an organisation is really hard.

Small service is simple. Big service is hard.

In that Milan restaurant service is a simple and elegant dance between service provider and customer.

The former gets paid with cash, loyalty and recognition (which converts into the goldmine of word of mouth). The latter gets food, drink, ambiance and their own sort of recognition. Done well, service creates this temporary bubble, where the customer is momentarily the centre of someone else's world - being looked after and comfortably moved from one state (hungry) to a better one (satiated). (NB Danny Meyer writes beautifully on the art of restaurant service in his book Setting the Table).

In my mind service is pretty close to holiness. And the church would agree. After all one goes "into service" when joining the church. All our thousand year institutions get this implicitly - church, army, government - all have selfless service hard-coded into their DNA. They're perhaps not always successfully at it, but their modus operandi is to "put aside one's own interests, in the interests of moving others from a lesser place to a better place". Church: damnation to salvation. Army: vulnerability to safety. You get the picture.

The problem is capitalism finds service really hard to do. Not impossible, just hard. More specifically, I think service quality has been going down the more mature and entrenched capitalist values become. And most of our organisations are in late-stage capitalism. It takes a maverick organisation to prove it's not impossible. More on that later.

(Ps. I'm not suggesting that businesses should give up the profit motive and selflessly prostrate themselves in service of the customer. You'd win lots of customers, but lose lots of shareholders.)

Some reasons why our organisations find service so hard

1. Misunderstanding what service is - most people in most organisations don't really know what service is. I once spent 4 hours in Boston City Library, one of the biggest libraries in the world, looking for a single book on service - it doesn't exist. Having spent years trying to pin it down, I can attest that it's a mercurial subject. One thing though is certain: service is not customer service. Customer service is where you end up when the service fails. When the thing you hired fails to deliver - a late delivery, cancelled flight, bill shock - you go to customer service. If your strategy is to improve customer service, you're starting from the wrong place. Customer service departments are cost centres, whereas true service businesses recognise that service is about revenue. It's about getting the customer to keep hiring you. (And by the way everyone IS in the service business. Think you're in the business of selling bottles of water? aka the product business. Wrong. You're in the business of removing thirst, which is a service. And by-the-by, there's a lot more value and innovation in delivering a great thirst-removal service. More in Robert Lusch and Stephen Vargo 's excellent book).
2. Size - the bigger you are, the harder it is to bring everything together in the moment for the customer. Data about the customer is invariably in the wrong place or the person with the expertise isn't available, which leaves the customer frustrated with a company that doesn't seem know who they are. The waiter has it all in his head. The business has to work hard to connect it. That costs money. Tom Peters writes a lot on this https://tompeters.com/
3. Fragmentation - the more you fragment the service chain, the less ownership anyone really takes for the service experience. Take the modern day clothes shopping experience. It starts on google, goes through a third-party hosted ecommerce platform, then via a payment provider, then through to a third-party delivery company, and finally through an outsourced and often offshores customer service aftersales function. Who in this chain is responsible for the ultimate service? Answer: none of them. No matter how good your "customer experience" strategy, or how well your service levels are written into your contracts, your service experience will lack integrity. I say this having spent 8 years working in outsourcers.
4. Lack of staff autonomy - it is very hard to codify service. To an extent you can write a manual and train people, but at some point it comes down to letting the employee do what's right for the customer in that moment. This is what is happening when hear people speak glowingly of companies that "go above and beyond." They are able to bend the rules and flex protocol selflessly, in the interests of the customer. If you don't trust your staff or don't give them autonomy, then these moments are much less likely to happen. Basically, staff need to be able to make up for all the other points. (Sidebar: most organisations are Theory X organisations (employees are lazy and carrots and sticks must be deployed to make them work), rather than Theory Y organisations (employees are intrinsically motivated to do good work and extrinsic carrots and sticks need to be deployed carefully so as not to disrupt that motivation) so it's no wonder really that autonomy is lacking. If you want to be inspired on the benefits of being a Theory Y organisation, read McGregor's work, or for a present day example of someone getting it right read about Timpsons Limited or read Employees First, Customer Second.
5. Short-term managerialism - service is born of culture, and culture takes years to build (and weeks to destroy). Yet business imperatives mean results are expected in months or quarters, and the average 3-year CEO tenure doesn't allow time for true investment. Instead we get short-term managerialism and technocratic solutions - aka solutions that stick to the dogmatic and tried-and-tested, which as the FT article shows, haven't worked and don't work. Add some new tech, reduce or ration access to customer service, offshore it. The outcome is what we all feel: that we the customer, are merely a product that the service provider is trying to process as cheaply and quickly as possible. Leaders will say they don't want this, and want to "put the customer at the heart of everything we do", but management theory doesn't allow it. Deming is really good on this. He had that crucial characteristics of great leaders - a growth mindset - and when that mindset travelled, particularly to Japan, he was able to reimagine so much and break so many management orthodoxies. Sadly his legacy has been pigeon-hold into the narrow field of "total quality management". I'd actually recommend this book by Scholtes for a neat summary. Bit dated, but still great. If you work in the public sector, then read Seddon.
6. Not enough engineers. Engineers? Related to services? Sounds weird right? But what do Larry Page, Jeff Bezos and Steve Jobs all have in common? They're engineers. Services are systems, and systems have their own intrinsic and often counter-intuitive behaviours. Upstream changes have complex, often unintended downstream implications. Apparently weird but highly predictable causal loops confuse everyone. Human behaviour gets warped by well intentioned targets. Engineers understand these system traits better than anyone. And it's my view that you can't make service work in service organisations without understanding them as systems. Alas, very few organisations have engineers as CEOs or in senior positions. Even fewer allow a systems thinking logic to challenge the command and control orthodox.

So what will happen with AI?

I made the leap from Service Design to Artificial Intelligence a few years ago. I saw how software ate services in the early 2000s. I saw AI coming and I definitely wanted to be part of it.

Arwen, the company I help run, uses AI to help clients get greater value out of social media. We started four years ago by solving the problem of toxic content. Our AI now processes hundreds of thousands of comments a day, automatically detecting and removing spam, hate and toxicity in real time.

Having tackled the bad stuff, clients started asking us about the good stuff. Their social media communities are full of people trying to find out about their brand, select the right product, ask how to use it, complain about problems with it - basically whatever customers have being doing in stores or on phone lines, is now happening in their social media communities.

The problem is that most social media teams are small, and the communities they're trying to support are massive (a ratio of 1 employee to 1,000,000 followers is not unusual).

So I now have skin in the game. Our business is at the cutting edge of a service revolution, where you'll be able to interact with an intelligent piece of software which will help you progress from problem to solution (eg unhappy and complaining, to happy and resolved. Looking for a solution, to owning the product. etc).

It's in my interest to avoid the problems software had eating service, and overcome them now that AI is eating service.

Phase 1: Bad AI Service

We're currently in phase 1 of the AI hype cycle - which is unsurprisingly a bit disappointing. It's full of crappy chatbots that don't even admit their machines, making a pigs ear of interaction, and leaving us all feeling poorly looked after.

These chatbots are powered by general-purpose language models, that don't require additional background knowledge. But service requires lots of background knowledge - about the customer, the context, the history, the organisation's rules etc. Good service chatbots require the next generation of AI

Phase 2: Better AI Service

At Arwen we're focused on the Slope of Enlightenment [tips ash from his gauloise and sips his espresso]. Arwen's artificial intelligence is able to automatically detect questions and complaints, read and segment them in real time, and generate responses in the client's brand voice.

Because service is a complex and knowledge-intensive tasks, the large language model-based system needs to access external knowledge sources to complete the task. This allows for more factual consistency, improves reliability of the generated responses, and helps to mitigate risks of "hallucination".

At Arwen we're playing with this new underlying technology (which Meta AI researchers have called Retrieval Augmented Generation (RAG)), to synthesise conversation from multiple data sources, and create factually correct, human-like, branded and compliant interactions. We call it Engage. It promises 1,000 conversations a minute.

The 'Must Get Rights' of Good AI Service

Good AI service is going to be hard. We will need to get these things right:

1. Be transparent. To avoid the uncanny valley, the technology needs to be honest about what it is. It should avoid pretending to be a human, as that just weirds people out, like those robot humans you see on TV (shudder). Paul Thurston wrote an interesting piece about labelling AI with kitemark. It needs to serve with humility, and dare I say it, personality.
2. Be connected. It needs to be connected to the right information about the customer and about the business, in order to fuse together what service designers call "moments of truth" - a touchpoint where your brand is realised in the mind (and heart) of your customer, through the delivery of value. If an AI doesn't have data, or it has the wrong data, it'll give crap service. Garbage In Garbage Out remains a perennial rule.
3. Be careful. Giving AI control of conversations on social media, for instance, will achieve something no-one has yet achieved - monetizing the enormous social media communities brands have invested so much in. Suddenly the social media "engagement mountain" can be managed rather than ignored. But we need to proceed carefully. Arwen Engage is in the co-pilot phase, recommending responses to a trained human who completes them. This AI + service innovation will be as dramatic or more than software + service. There will be a real drive to unleash it on the world and get to scale fast so the competitive edge can be bagged. But we will need to manage that transition carefully. After all, AI is built by fallible emotional leaky bags of flesh and bone (I love the Arwen team), so it will also be leaky.

Undoubtedly AI eating service brings big ethical and social questions. People are right to be anxious about it. But I'm excited by what's to come.

AI eating service is going to bring about some pretty profound changes. AI eating social alone will enable things that were previously impossible - like answering questions posted on a social ad, determining purchase intent, and helping someone find and buy the product.

I mean, imagine the day when you can tell the CFO that the social media team is now a revenue generator, rather than a cost centre!

It's closer than you think.