Everyone is worried about the urgency to use Artificial Intelligence in their businesses, professions, and even their personal lives.
The FOMO (Fear Of Missing Out) is greater than ever.
And the main question we hear nowadays is: HOW DO I use Artificial Intelligence in my business?
But the real one you should be asking yourself is: SHOULD I use Artificial Intelligence in my business?
To answer it we need to first understand a few things:
How my business makes money today?
Can I make more money in my business by using Artificial Intelligence?
Will the profits from using Artificial Intelligence in my business be greater enough than the costs of implementing it?
How long will it take to start getting results?
Only then we can decide if we can truly profit from using Artificial Intelligence, and not only spend money following a fad.
In Today’s Issue:
Understanding how your business makes money today (a.k.a. “Your Profit Model”)
The Current State of AI
How AI Changes Things?
Should You Implement AI in Your Business?
Profit Models for the Age of AI
Understanding How Your Business Makes Money Today (a.k.a. “Your Profit Model”)
You might be thinking: “But I KNOW how my business makes money! I found a market need, I’ve developed/procured a product I can sell at a profit big enough to be worth the effort, and I’ve been doing it for some time”.
And, yeah, that’s a very general view, but not enough for our analysis.
You see, both Ferrari and Toyota are sell cars and are profitable. Right?
But one of them has a lot of factories and sells millions of units, all over the world.
While the other has few very specialized factories, and sell a few thousand units a year at most.
Let’s not even talk about the price difference between any Toyota and a Ferrari.
They sell differently, they produce differently, they have very different workforces, and they provide very different associated services (post sales, etc).
So, it would be logical to think any attempt to apply some new technology, like Artificial Intelligence, will be very different too.
Their Profit Models are different.
So their revamped Profit Models, ready for the Age of AI, will for sure look different too.
An excellent read on this topic is Adrian J. Slywotzky’s book “The Art of Profitability”1.
It is an introductory book on how different businesses have different ways of making money, and even businesses that look very similar can also have wildly different ways of making money.
So let’s take one of his proposed Profit Models as an example: The After-Sale Profit Model.
You sell “big stuff”: houses, cars, expensive equipment, etc.
It takes you a lot more than a few calls to get the sale.
It requires many calls, emails, visits, decisions from multiple people in your customer’s organization, etc.
This can take weeks, even months.
It also requires a very specialized sales team, with certain specific skills, and a corresponding commissions structure to incentivize them.
So, once you get a sale, it’s a big celebration.
Lots of money after months of hard work.
Now compare it to an organization that sells small ticket.
Probably you have thousands of customers per month, or even per day.
Your sales force looks very different, their incentives are completely different, and you focus a lot on logistics because your margins are very thin, and every penny counts.
An organization that sells big ticket usually focuses on getting the big sales, but misses the sea of after-sale opportunities that open at the moment the big sale is achieved.
For example, once you sell a car, you can also sell customized products for it, insurance, infant car seats, safety equipment (like portable jump starters and compressors), etc.
It works the same for houses: appliances, furnishings, insurance, etc.
But you will see most businesses waste those extra opportunities to make money because they don’t have an organization capable of handling the small ticket sales that come right after the big ticket one.
So, in a pre-AI world, you would realize this opportunity, and create a parallel sales organization capable of handling those small thingies once the big fish sales team finished the big sale and moved immediately after their next opportunity.
But now we live in the Age of AI.
And this completely rewrites the playbook.
The Current State of AI
First we need to understand where we are regarding what we call now “Artificial Intelligence”.
Skynet from Terminator, the Agents from The Matrix, Data from Star Trek, etc. are what we would call today Artificial General Intelligence (AGI).
That is a level of intelligence comparable to human intelligence, with not only the same cognitive capabilities, but also with agency (its own will to do things), a world model (understanding how and why things are like they are), etc.
Obviously we are not there yet.
By far.
Current Artificial Intelligence can be classified under what we call Weak AI, in comparison to Strong AI or AGI.
It simulates human intelligence by parroting back probabilistically generated answers, that are pre-trained into neural networks by consuming almost-absurd amounts of computing power.
Basically it answers what it calculates as the response that most probably matches what has been asked, based on the data it has been previously fed.
It doesn’t know if it is the right answer.
It only gives you what it looks like the right answer, when you take into account all the data it has on the topic.
So, by its own nature, it is prone to fail.
In response to that, we have developed guardrails, so it doesn’t wander away from topics where it has trustworthy data, and computational techniques that allow us to make sure they answer from up-to-date curated pools of knowledge (Retrieval Augmented Generation, or RAG), etc.
Does it still fail?
Yes.
But not so catastrophically, and not so often as before.
So it is virtually impossible to make it 100% accurate and trustworthy (something we have been accustomed to expect from computers) by the very nature of its inner workings.
Nevertheless, by using it carefully and in the right scenarios, it has proven an extremely powerful tool and force multiplier for almost every domain where knowledge is created and manipulated.
As long as we have humans validating that our Weak AIs are not hallucinating (the name the industry gave to the tendency of AI of producing wrong answers that look like the right ones), we can greatly profit from them.
This is called the Human In The Loop (HITL) pattern.
But we weren’t happy enough with a powerful tool sitting in our desk depending on us to be operated.
We wanted to make it autonomous, so it could work while we sleep.
And this is how Artificial Intelligence Agents came to be.
While still prone to error, albeit powerful, AI models could be encapsulated in what we call a Cybernetic Feedback Loop: an endless cycle of looking around, performing a requested task, obtaining a result, checking if it was the desired one, adjusting itself to do it better the next time, and again looking around for input and resources.
So now we can have dozens, even hundreds, or thousands of AI Agents, trained to fulfill specialized tasks (like Finance, Operations Optimization, Software Development, etc.), working towards our goals.
24/7.
Without sleep, without sick days.
Tempting, right?
But don’t forget the weak in Weak AI.
Those AI Agents can do marvelous things, faster and cheaper, but can also commit marvelous mistakes, because they aren’t really intelligent.
As mentioned before, current AI lacks a world model, so what would be obvious for any human, it isn’t for them.
We, as humans (and even animals), know that something called gravity exists (even before giving it a name, or knowing the physics behind it).
Some time, before we were 1 year old, we fell to the ground while trying to stand up for the first time, or rolling off a coach, etc.
This meant an immediate OUCH!
And OUCH teaches us, better than any book or PowerPoint.
We learn that when you drop something, it doesn’t float away, but it falls to the ground.
We have a world model.
Current AI lacks that kind of knowledge.
So, if you insist enough, you might even convince it that pigs can fly.
Or it can come to similarly ridiculous conclusions if fed with tainted data.
So, you should treat your very technically competent AI Agents like you would a freshly graduated intern: give it increasingly complex tasks, and expect mistakes.
This means you strictly follow the Human-in-the-Loop pattern mentioned before: always check what your AI Agents produce before giving it the go.
You should treat your very technically competent AI Agents like you would a freshly graduated intern: give it increasingly complex tasks, and expect mistakes.
How AI Changes Things?
Now let’s go back to our After-Sale Profit Model example.
You have a business that sells big ticket: houses, cars, expensive appliances, etc.
Your organization is tailored to effectively acquire clients willing to spend big money on your products.
But then, all the small ticket items you could have sold as part of the big sale, and that can also become a recurrent cashflow (think of razor cartridges you recurrently purchase for a more expensive razor handle) are left unattended, and someone else profits from them.
Your specialized big ticket sales hunters couldn’t care less for small ticket but recurrent cashflow items, so you would need a whole new sales organization to take advantage of this opportunity.
In a pre-AI world, that’s precisely what you would have done: analyze the potential cashflow and profits you are leaving on the table, and decide if you want to expand your operation to also sell those new smaller items to your existing customers.
Now you can do the same with a smaller organizational footprint by using AI Agents.
A typical human organization would need two main components: the ability to sell and the ability to fulfill the sales.
It’s AI counterpart will need the same, with AI Agents split between the corresponding roles and necessary functions.
To be able to sell, you would need the following line-up:
A Marketing Research Agent, trained to find out what small ticket products would highly increase the convenience of the recent big ticket purchase, research the market prices and how much customers would be willing to pay, how much it would cost to fulfill the sale, and potential profits. Then it would select the best ones in terms of both profits and customer convenience.
A Sales Preparation Agent (or Qualification), trained to traverse the company’s CRM to create a list of potential customers, the products that better match their needs, a sales pitch tailored for them, and preparing the CRM with the required info for the sale itself.
A “Salesman” Agent, trained to perform the actual initial outreach (sending emails, text messaging, or even AI-enabled voice calls). Like any salesman, it would follow the already prepared scripts, follow up through different channels, and try to close the sale. If human intervention is needed, it should act as a router to an Human Agent.
An Analytics Agent, that collects all the info regarding the sales, and feeds back the Marketing Research Agent, so it can decide if the expected profit is being achieved, the products are being well received, and if the sales process itself requires any fine tuning.
But once the sale is done, you need to fulfill it.
That part can also be handled by specialized AI Agents:
A Resource Planning Agent, working closely with the AI agents from the sales organization, it would plan how many of each product are necessary to keep the customers satisfied, while minimizing inventory, or even better, working in a just-in-time fashion.
A Procurement Agent, trained to find the best sources for each product, even negotiating deals and either having a pre-approved budget for certain purchases, or preparing everything for a human employee to quickly review it and approve the purchase.
A Fulfillment and Delivery Agent, specialized on putting the products on their way, and then tracking their delivery status, solving any issue that might arise by itself, or notifying a human to take charge if it cannot deal with a situation.
But small ticket sales can be prone to defects and require a proper remediation channel.
Specialized AI Agents can be used for Customer Satisfaction and Customer Experience roles too, backed by a few human Customer Experience Specialists, becoming a huge force multiplier.
A Customer Service AI Agent should be capable of solving 80-90% of the support requests, having access to the company’s knowledge base and sales and products information. Whenever it is unable to solve the problem, it should route the case to a human agent.
A Knowledge Base Agent, that will be used by the Customer Service Human Agents to solve the issues presented by the customers. Let’s face it, Customer Service has one of the highest turnover rates in any company, and retraining new agents is both lengthy and costly. Having an internal AI “expert” can greatly reduce the need of upfront training, allowing the Human Agents to learn along the way, while ensuring a high resolution and satisfaction rate.
A Remediation Agent would be responsible of coordinating any defective products Returns and Replacements, and if deemed necessary, any Remediation (small gifts as compensation for the troubles).
And finally, a Customer Experience Analytics Agent would keep track of every Customer Experience metric collected during the presale, sale, and after-sale, to optimize Customer Satisfaction, within the boundaries of profitability.
Should You Implement AI in Your Business?
So, as you can see, it is perfectly possible to leverage AI Agents to create a completely different sales organization that can take care of a new type of market and products.
But you do so without forgetting what we already discussed: always keep experienced humans in the loop, exactly as you would if you were delegating work to a new and inexperienced intern.
This means that the answer to the question boils down to two critical factors you should evaluate:
If the Profit Model can be enhanced by using AI, like we did in the After-Sale Profit Model example.
If the application of AI can be performed profitably.
The second point here is often overlooked during the decision process by business leaders hit by the FOMO (Fear Of Missing Out) madness.
They jump out into the proverbial empty pool, hoping for it to magically fill with water while they fall into the all predictable disaster.
Implementing AI in an organization can require an important capital investment, followed up by an also important recurring operational expenditure.
Not only you need to acquire subscriptions to different AI vendors (depending on every specialization you might need), but also train your models with your proprietary data, so it behaves as you would expect.
You will need external expert consulting to execute the implementation projects, and also an in-house team capable of receiving all that knowledge and keep it going long after the consultants are gone and the projects are finished.
AI deployment not only means tech, but also an important Cultural Change, that must be managed or your own staff might end up sabotaging you efforts, out of fear.
Managing multiple projects, and tracking your objectives through the year require some kind of Agile Management by Objectives capability in the organization, or you risk spending big money and failing to achieve anything in the end.
And once it is all done, you’ll be left with the recurring costs of either AI capacity you rent from big vendors (in the form of tokens), or with the costs of maintaining your own AI infrastructure, if you decided to go with an in-house solution.
So, once you put it all together, if the projected cashflow still provides reasonable profits, or gives you some strategic advantage that justifies it, you can green light the whole thing.
Profit Models for the Age of AI
In this article we focused on one of the many possible ways of applying AI to a classical After-Sale Profit Model to update it for the Age of AI.
The same can be done with most of the existing Profit Models, not only in Slywotzky’s book, but everywhere you look hard enough.
And often in more than one way, so there is a lot of room for creativity.
I will dedicate many of my future articles to explain existing Profit Models, and propose how AI and other Exponential Technologies can be used to make them even better performing, and put more money in your pocket at the end of the day.
So, if you don’t want to miss my next article, just subscribe for free to this publication, and you will receive it directly in your inbox every time.
Conclusion
Thank you for reading my publication, and if you consider it helpful or inspiring, please share it with your friends and coworkers. I write weekly about Technology, Business and Customer Experience, which brings me lately to write a lot also about Artificial Intelligence, because it is permeating everything. Don’t hesitate in subscribing for free to this publication, so you can keep informed on this topic and all the related things I publish here.
As usual, any comments and suggestions you may have, please leave them in the comments area. Let’s start a nice discussion!
When you are ready, here is how I can help you:
By subscribing to my publication and reading this article you have already taken the first steps into redesigning your business to succeed during the Age of AI, .
But if you want to go further, here are several options:
If you want to start right away, you can join me in my 3-day Future-Proof Entrepreneur Challenge, where you will learn how to analyze your Profit Model (how your business makes money today), look around for the Exponential Technologies you will need to make it work in the Age of AI, and finally create a plan to make it happen.
You can check my FREE guide “7 Profit Models to Succeed in the Age of AI”.
And also my FREE guide “7 New Technologies that can wreck your Business like Netflix wrecked Blockbuster”.
The Entrepreneuring AI is free today. But if you enjoyed this post, you can let me know that my content is valuable by fueling my coffee addiction, so I can keep producing it for free 😸☕️
References
Slywotzky, A. J. (2002). The Art of Profitability. Warner Business Books.
The shift from 'how do I use AI' to 'should I' is the right one, and most skip it. Treating agents like fresh graduates is honest: useful, fast, and wrong in ways you would not expect. The recurring token cost is the part business owners miss in the math. Did you model that cost against the agent lineup in the post?