How AI Sales Actually Works
By Jeff Billings
The core question every business must answer is: How to create a virtuous cycle of value? Every other part devolves from that question being correctly answered.
To optimize and reinforce the virtuous cycle, AI needs to be integrated into the pain points and leverage points of that cycle. The AI deployed must reduce the pain and increase gains… But how does that actually work?
Customer Acquisition Cost (CAC)
CAC is the cost to acquiring a new customer relationship. Seems obvious, right? But the ability to acquire a customer is… hard. The factors in this process make it practically impossible to achieve a complete detailed diagram with all the interdependencies.
We can simplify the idea to this equation:
Customer Value (CV) – Customer Acquisition Cost (CAC) = Additional Revenue (AR)
AI can only optimize measurable systems. Marketing in customer acquisition is often a creative, almost mystical artform with some trappings of countability. AI can’t help in CAC since the process is fluid, and poorly measured. To use AI to improve this key business metric, we must take the traffic generated by marketing and improve the acquisition rate, thus converting visitors to customers more efficiently.
Above are actual results where AI interacted with only 22% of site traffic. It drove a 29% increase in AR. The leverage provided by AI sales has a clear benefit to the individual customer value and acquisition rate. Just like a great salesperson, AI can understand what the customer wants.
Awareness, Interest, Desire, and Action (AIDA)
Marketing is centered on the AIDA model – Awareness, Interest, Desire, and Action. Let’s examine how users experience company communications right down to the completion of a sale to show where AI can create leverage over that cost expression. We use the term cost expression since all the money spent in these efforts by the site owner is cost of sale.
Awareness
Generating awareness of your site, brand or product through advertising, earned media or SEO optimization is typically a static effort, repeated with accompanying costs. For example, brand advertising is essential to establish a brand value promise in the public’s mind and is spent when a sellable product is attached to the brand. SEO on the other hand may look for streams of organic traffic where a feature of a product aligns with a buyer’s desire. Both can be aided by statistics but not Artificial Intelligence.
The best use of AI is understanding details of landing pages once a user has proven awareness through a pageview. AI can observe referring pages, URL variables, time of day or other latent variables to start building an understanding of eachuser. Beyond those details, AI on a bounded domain of use cannot influence awareness.
Interest
Once a user is aware of a domain of communication, the AI can adaptively interact with the user. The AI will work to achieve success – the success event is defined by the domain (the bounded knowledge – a website, library, or knowledge base) where the answer the user desires exists… and desire is the next step in successful AIDA.
How does the AI drive interest? It learns how to understand each person and then how to communicate with them to help them succeed and find what they are looking for. This isn’t just UX design, it is real-time communication based on the needs of each user. AI learns how to read the mood and interest of the user, like a great salesperson. Then it communicates what they want in a manner that moves them to confirm the user’s desire by adding to cart.
Desire
When users express their desire, the AI gets a clear signal that it is aligned to their thinking. We call this highly accurate measurement, Directly Attributable Revenue (DAR) telling us that the user saw a communication from the AI, interreacted with the AI in a single session and completed the desired action, all in one session. This is as close to unambiguous proof of direct AI effect as can be measured.
There are two critical factors involved in creating an effective AI sales event.
- User to AI interaction ratio. If 1000 users visit a domain and the AI only interacts with 220 of them… then 780 could never be successfully affected. This discount rate is often driven by UX designers assuming the AI interaction components are in the way. What is often missed is that the AI components need to be in the way for optimal interaction.
- Traffic rates versus SKUs. For AI to learn it must see what worked on a domain of use. The best AI systems can learn with a limited number of examples, but most need hundreds of success events to learn from. Let us use this example… 2000 SKUs with a daily visitor rate of 2000 users per day with a standard conversion rate of 2.3% of visitors, and 100 SKUs are top sellers.
USERS-PER-DAY ÷ TOP-PRODUCTS = TOP-PRODUCT-INTERACTION-DAILY TOP-PRODUCT-INTERACTION-DAILY * CONVERSION-RATE = SUCCESS-EVENTS SUCCESS-EVENTS * AI-TRAINING-RATE = LEARNED-BEHAVIOR
If the AI required 300 examples of success to learn, it would take 653 days (about 2 years) to learn the domain for the top 100 selling SKUs. Waiting that long to see if it works isn’t really an option. Look for AI that has an extremely small success observation learning rate. For example, the same case with an AI that can learn from just 17 examples could prove the AI is working in 8 days.
Captured Cash
Cash flow failures often lead to the downfall of businesses. This is an area where AI can help reduce risk. Consider the 1900 slow-moving SKUs which are slow movers on our example website. Depending on the Retail Position of a store, such as Biggest or Easiest, where other SKUs play a crucial role in customer conversion; these SKUs can capture cash and pose a risk of loss. To mitigate the risk, each shopper needs to add the appropriate complementary items to a cart– increasing the overall revenue and decreasing the captured cash held in inventory. Many AI systems lack the ability to do this. For AI to do this effectively, it must:
- Have enough visitor interaction rate with the rest of the SKUs in the inventory.
- Understand visitor preferences quickly enough to offer the underperforming SKUs.
When an AI can both learn the “back catalog” and determine when to sell it, revenue soars and the otherwise captured cash starts flowing. Both are essential to business health and growth.