- “Small business” and “AI” is a killer combination. Why AI— that’s understandable 😉 But why small business? Because small business— that’s 99% of all companies in any country.
- Moreover, 80% of small businesses don’t have any employees at all! And another 16% have fewer than 20. So, such businesses have a bunch of urgent tasks that simply no one is available to do
- No one— except AI, which can be taught to handle these tasks. It turns out that such businesses will eagerly pay for such an AI platform like this:
Project essence
Technically, Twin is an AI tool that “turns text into actions.” Essentially, Twin is a platform for automating small businesses.
The startup’s main slogan is “don’t do the same thing twice.” The platform’s task is to ensure that owners of small businesses and their few employees don’t waste time repeatedly performing sequences of the same actions on the computer to complete a work task.
The AI machine should learn to perform this sequence of actions when a person does it for the first time—and then execute it on demand.
For example, when a company owner hires a new employee, Twin can prepare a contract for them from a standard template, upload it to Docusign, and send it to the employee for signing, order a laptop for them on Amazon according to standard specifications, and register them in Google Workspace, Notion, and other platforms used in the company.
Similar tasks exist in large numbers not only in HR but also in sales, development, and operations.
As a result, Twin becomes an all-around assistant in a small company, to whom various routine tasks can be delegated.
At the same time, the founders claim that they first tried for a long time to teach the AI machine of the platform to perform such tasks using text instructions—but in general, this was not successful enough. Then they decided to approach it from a different angle—from the side of computer vision, and this approach turned out to be more successful.
The platform’s AI machine monitors what happens on the screen when a person performs the necessary sequence of actions and also listens to explanations via the microphone—thus, it learns to perform the necessary action on its own.
In a broad sense, this is exactly how we teach employees about work processes. If we simply give them text instructions, they won’t figure it out themselves anyway, and we’ll still have to show and tell them afterward 😉
Currently, the startup is recruiting its first “partners”—actually, future clients. Together with them, they will automate the most common tasks to populate the platform’s library—so that other clients won’t have to train the AI machine to perform typical tasks.
Twin hasn’t even launched yet, but it has already raised its first $3 million in pre-seed investments to develop the platform and populate the library of typical tasks.
What’s interesting
Most startups aiming to capitalize on AI technologies are involved in content generation tasks—such as websites, emails, articles, advertising banners, blog posts, and social media posts.
“Turning into actions” is another task, although the same technologies are used to solve it. However, as we see in the case of Twin, this task turns out to be more complex—and here you need to figure out how to implement it so that it works reliably in various situations.
Moreover, to achieve acceptable efficiency, it is necessary to automate not just a single action but a complex task that must be performed entirely on autopilot. In this case, creating instructions for the automatic execution of such a complex task should not be more difficult and time-consuming than performing the task manually 😉
From this perspective, today’s Twin resembles the startup AskToSell, which I wrote about in September. This startup has created a platform that can automatically make small sales—for which it is not profitable to use live salespeople.
Their platform’s AI machine selects contacts from the lead database, writes emails to them, responds to replies, makes purchase offers, negotiates prices—and involves the business owner only when it’s time to accept or reject the sales conditions negotiated by the AI machine.
AskToSell is a European startup that is just launching its platform, so it has attracted only €100,000 in investments so far.
The target audience of AskToSell is also small businesses, most of which have small sales volumes. Therefore, it’s not only unprofitable for them to hire live salespeople but also impractical to engage in such sales in essence.
Because 80% of small businesses have no employees at all! 16% have from 1 to 19 employees. And only the remaining 4% have more than 20 employees who can handle something 😉
Moreover, most companies in any country are small businesses. For example, in the USA, small businesses make up 99% of all companies. And this, mind you, is 33.2 million companies in just one country!
So, Twin’s focus on small businesses seems quite justified.
Twin’s platform development plan consists of 3 stages:
The first stage is to create a library of typical tasks, which account for 20% of all possible tasks but take up 80% of the time. The second stage is to provide companies with an AI machine that will gradually automate the remaining 80% of simple tasks. The third stage is to use the accumulated data and experience of automating simple tasks to develop the AI machine towards automating more complex and very complex tasks.
If you look closely, the file title of the plan on the screenshot is “twin-master-plan.” The “master plan” format was popularized by Elon Musk, who published a “secret master plan” in 2006, outlining three stages of Tesla’s development:
Build a sports car. Use the money earned from its sales to build an affordable car. Use the money earned from its sales to build an even more affordable car.
Lately, I’ve come across several startups that also clearly show their “secret” master plans.
Startup Siro is developing an AI coach for salespeople. Their master plan:
In the first stage, the service should record salespeople’s conversations, analyze them, identify weak points, and send them to the sales department manager to provide feedback. In the second stage, Siro should send salespeople the weak points of their conversations along with examples of strong salespeople’s responses—so they can learn from their own mistakes and from the best examples related to the specific mistakes they make. The task of the third stage is for the Siro AI machine to be able to formulate general principles from the best practices and immediately suggest ways for the salesperson to apply them to the mistakes detected.
Startup Evolv AI is creating an AI platform for website optimization. Their master plan:
An AI machine that allows making changes to the website design without involving programmers. An AI machine that can automatically test different design options and provide a comparative table of their effectiveness. An AI machine capable of suggesting changes to website owners based on the collected statistics from all platform clients for possible conversion improvement.
Startup Durable is creating an AI machine capable of creating websites for small businesses. Their master plan:
The first stage is a builder for simple websites that can attract clients wishing to solve the simplest and most obvious tasks of their business. The second stage is the creation of AI tools that allow automating individual business processes related to having a website: CRM for processing requests received through the website, writing articles for the blog, and so on. The third stage is a “complete package” of AI tools that allow automating the entire business, including marketing, sales organization, financial accounting, taxes, and everything else.
Where to Run
First of all, regardless of what your startup is or will be doing, create a master plan for its development, which should also consist of 3 steps 😉
The goal of such a plan is not to endlessly stretch the process of creating the perfect final product but to move towards it systematically:
- Quickly create the first product with minimal functionality to validate the basic concept and start making money from its sale.
- With the earned money and using accumulated experience, expand the functionality to a new qualitative level—again, creating a full-fledged product that can be sold.
- Only at the third stage, implement what was originally conceived. Naturally, with the changes that will be caused by the real experience of using products created at previous stages. Or do the same trick—but with expanding the target audience, consistently earning money by creating new products for a wider audience.
The general possible direction of movement, however, is the creation of AI platforms for small businesses.
The point is that there are a lot of such businesses, and they need specific tools. On one hand, these tools should be very easy to learn and use, but on the other hand, they should almost completely take over the entire cycle of performing existing tasks so that the small business owner doesn’t have to hire separate employees for this and doesn’t have to do it themselves.
Options for such tools may not even be very common. For example, the British startup ProperPlan is creating an AI machine that will provide small business owners with ready-made action plans to achieve their business goals. The platform is still under development, but they have already raised £300,000 in investments for it.
And what other common tasks or processes of small businesses can be automated with AI? How to make a tool easy to learn and use for this purpose? What technical difficulties will you encounter? How can these difficulties be solved?
And finally—how to break down the process of creating the “ideal” product into three stages, each of which will result in a self-sufficient product that can be sold? 😉