The Idea In 60 Seconds:

  • When we all use AI Assistants there will be implications for the businesses we deal with.
  • Those AI Assistants need to be able to do the same things human visitors do on your site.
  • Many companies have spent the last 20 years making their services available online.
  • One way to deconstruct a website is to look at it from the point of view of 3 specific intents.
  • 1) Evaluate – Where visitors come to your site to find out about your product. (Presales.)
  • 2) Buy – Where they have established enough information that they can now buy. (Sales.)
  • 3) Self Service – Where they have bought and are now maintaining their product (Postsales.)
  • Some industries more than others are going to find it hard to adapt to the change.

Let’s Say AI Assistants Are Visiting Your Site On Behalf Of A User – What Do Businesses Need to Do To Prepare?

The previous article dealt with the likely sort of effects businesses will face if the world starts using AI Assistants / Agents. This article tries to get a bit more concrete about what businesses might like to do to prepare for and respond to a situation in which sizable chunk of their visitors are AI Assistants rather than people.

I’ve worked on upgrading a number of websites. It’s a big job. One helpful way to break websites down was to consider 3 intents people have when they visit a site : ‘Evaluate, Buy, Self Service.’ I’ll work through each 3, considering :

  • What facilities the AI Assistant will need when it gets to a site with one of those intents.
  • Give an example of the sort of operation the AI Assistant might undertake.
  • Suggest improvements that could be made to the current state for each of those intents.

Evaluate – When A Visitor Is Evaluating A Product

When an AI Assistant, acting on behalf of a user, is visiting your site, to evaluate a product, they are looking for information. In these circumstances, the AI Assistant will require the facility to acquire information from the website, filter using user / Assistant selected criteria to find the right product and then compare a different provider’s (here, your) products to pick a winner. For example, someone might want a phone plan, on the latest iPhone which has at least 30 GB of mobile data each month.

As an example, imagine the AI Assistant goes to Vodafone, Telstra and Optus’ websites, filters by iPhone 15, scans for plans which have at least 30GB per month, notes any special criteria (e.g. you must buy by the 15th or this promotion will end / This iPhone is a returned unit and has been refurbished.) Then the AI Assistant compares the same offers from Optus and Telstra and provides a summary to their user. All of the information relating to presales investigation needs to be made available to the AI Assistant when they get to your site.

What Do Websites Need To Do? Probably Not A Lot For ‘Evaluate’

I believe ‘Evaluate’ is well provided for on existing websites. Most websites have a thorough information depository. Evaluate is presales activity and whenever sales is involved, a lot of resource is applied to doing what is required to sell things online.

What might change when a chatbot comes to your website instead of a user? The only difference I can see is the chatbot might require an answer to more long tail (less frequently requested) questions. Those can be captured and the library extended.

Tie in conversation with visual. People are used to scanning large pages of text and picking out the information they need. At some point, when evaluating, they meet the emotional and practical thresholds for performing the purchase. Maybe the user wants a Product Data sheet to print off at home? Or a comparison matrix. Having a list of content assets the AI Assistant can provide to the user (or request if you don’t have them) seems sensible. Like I say, websites are pretty well set up for ‘Evaluate’.

What About When an AI Assistant Wants To ‘Buy’ For Its User?

When an AI Assistant comes to your site to buy something, they need to be able to complete the purchase on behalf of the user. For now, it’s hard to imagine AI Assistants being given ‘Power of attorney’ – that is to undertake legally binding activities on behalf of their user. It’s far more likely that businesses will have to develop a facility for the AI Assistant to populate many of the fields required to buy the item but that the actual user themselves will sign the contract, fill in the CVV number and accept the terms and conditions.

To continue the example above, to buy a phone, the user tasks the Assistant to go to the website, enter their personal information including credit card details. The Assistant might need to wait while a credit check is undertaken. Then, behind the scenes, an account is opened in the customer’s name. 2 SKUs are ordered a phone and a SIM provisioned with a plan. Those things are collated in a warehouse and shipped to the person. I think every step of that process has to be made available through a well documented API (Application Programming Interface – a concierge in one software platform that helps visiting software find that it needs) so AI Assistants can access it. But the user themselves will have to roll their sleeves up and undertake the legal and financial liability.

Developing an experience which minimizes the hassle for the user and passes inconvenient activities like ‘waiting’ (for example, while a credit check is completed) should be passed to the user.

My experience of eCommerce is that there are many things which can throw a spanner in the works at this time. Users have all sorts of queries about the checkout process. Oh. Maybe I can find a promo code. Does this renew every 28 days or calendar month. Answers to all of those questions need to be readily available for the user / Assistant so the sales train is not derailed.

In terms of improvement, Assistants need to be able to buy the whole thing, as much as possible, online. In telco, with an eSIM, the ideal is that the service can be provisioned when the device is purchased. This is a good example. The process to provision (set up) an eSIM is still clunky and is still not available to smaller phone companies at the moment (MVNOs.) It’s a good example of the sort of thing businesses need to anticipate and enable before AI Assistants become commonplace.

Similarly, Self Service. What Happens When An Assistant Tries To Maintain an Existing Product?

Most of the sorts of jobs people want to do to maintain or upgrade their products are available online.  Big companies have spent the last 15 years trying to lower their call center costs by making customer online self service available and free. Enabling these existing services as eServices to AI Assistants is a job but it’s much easier than having to create the service in the first place.

Again, using telco as an example : Say the phone has now arrived at the user’s house and he’s been using it for a few weeks. He wants to move from a plan with a 30GB plan to a 50GB plan. That service is almost certainly available by now from a phone company. The Assistant can go to the website and login and perform the task online. And / or the company can make those services available as eServices through an API. The latter sounds cleaner and less accident prone. Again, there may be some legal requirements that the end user has to accept and the Assistant can’t do for him/her.

In terms of improvements to self service activities, my personal point of view is that the focus should be on mining the most frequently performed self service activities and data mine them. The best product is one that never needs servicing and using some of the new facilities Generative AI provides could take businesses in to a more proactive self service world which their customers would much prefer.

Some Industries which might find it Tough To Turn On eServices

So far, in this article, I’ve covered Evaluate, Buy and Self Service; what facilities an AI Assistant would need to have present when they got to a website with one of those intents and potential improvements businesses could make. However, some industries are going to find it harder than others to enable their services online. Businesses with intangible products (services, especially those delivered directly by a person), businesses with strict security / privacy requirements, small businesses without the same resources as bigger companies and those with intricate process elements – some aspects of Industrial Manufacturing, for example, might well find it harder:

  • Healthcare: The healthcare industry faces unique challenges due to the intangible nature of its services and the need for personalized care. It’s an industry where expertise and quality of care often outweigh cost considerations. Additionally, smaller, localized medical practices may not be able to leverage economies of scale as effectively as larger organizations, making it difficult to implement eServices on their websites widely​​.
  • Service Industries with Intensive Security Needs: Industries that handle sensitive financial, health, or identity data, such as banking and healthcare, face significant challenges in maintaining the security of their eServices. It seems likely their customer experiences will need to be closely mapped and the elements which the Assistant can perform on behalf of the user carefully considered such that privacy is maintained.
  • Small Businesses with Limited Resources: Small businesses often struggle with the affordability of making their websites and eServices accessible. The cost of accessibility solutions, whether automation tools or manual audits and modifications, can be prohibitive for smaller companies. This challenge is compounded by the vast number of websites and the dynamic nature of the internet, making a manual-only approach to accessibility impractical​​.
  • Industries Requiring High Levels Of Physical Interaction: Industries that rely heavily on in-person interaction, like hospitality, may find it challenging to replicate the physical experience through digital services.
  • Industries with Complex Service Delivery: Some industries have intricate and specialized service delivery processes that are difficult to replicate or streamline through eServices. For example, sectors that require hands-on, personalized attention or involve intricate procedures (like certain manufacturing sectors or specialized repair services) might find it challenging to transition these services to an online format.

We Need To Re-imagine Every Step of Customer Experience

When we all have AI Assistants working on our behalf, we’re going to expect different things from the websites we and our agents visit. Carefully planning out customer journeys, bearing in mind the legal and commercial obligations that the user can only do themselves will be a job for every website over the 2 years ahead.

Some of the issues relating to privacy, security and the abilities of AI Assistants are so significant (can the AI Assistant oblige you to a contract for a service? Which services? When?) that regulation might be required. Which is fine but may be made more difficult by the speed with which these changes are going to occur, the next and final article.