Watch your mouse and keyboard complete a preset task on your own computer

1. It’s not just one click

Today you need to write prompts by hand, every time. Tallyfy presets instructions, inputs, outputs, guardrails and much more, since we only focus on repeatable tasks. This means everything is ready for anyone at all to just click Do It For Me and instantly watch their computer come to life. This capability is critical for adoption.

2. Adoption is a huge problem

People find it difficult to use AI beyond ad-hoc, random tasks. The biggest and most powerful opportunity for AI is the ability to use it for preset, preknown tasks. Tallyfy is the critical plug (Tracking AI for workflow) that enables everyday use of the AI socket.

3. Continuous improvement is hard

Everyone knows that AI prompts or guardrails are never right or perfect the first time. Tallyfy offers the simplest possible feedback solution from users to moderators on how to improve AI for a task, add guardrails or propose improvements.

Computer AI is the enabling socket of technology supplied by LLM’s like Anthropic, OpenAI, etc. It has the core capabilites of seeing what’s on your screen, and controlling your mouse and keyboard. However, just like a socket – it’s of no use unless you actually plug in an appliance to make use of the power.

Trackable AI is Tallyfy. It’s the plug or workflow platform you need to use Computer AI in a practical way. It covers huge gaps you need to fill in these areas of a task.

• What to do – step-by-step instructions
• What not to do – guardrails
• Pass through information the AI requires – input variables
• Collect output from the AI in structured format – form fields
• Skip tasks that don’t need to be done – if-this-then-that conditional logic
• Report a problem if the AI fails – blocking comments
• Let people suggest better instructions or guardrails – improvement comments
• … and more

A small preview of just one thing you need for dynamic data (input variables) – is shown below.

Computer AI requires Trackable AI for business users to do repeatable tasks within a larger workflow. Going back further – before Computer AI, many of us were familiar with Text AI like ChatGPT, Claude, etc. Text AI works with an input of text or images – and an output of text or images. However, Computer AI goes a lot further – it uses that, while also controlling your mouse and keyboard to get a small, defined task completed on your computer.

The ideal use cases for Trackable AI + Computer AI are repeatable tasks within repeatable workflows. A good automation candidate is a task that’s that’s short, administrative, mistake-sensitive or difficult to complete successfully – especially for newbies. Such automation candidates exist everywhere in the long-tail of business processes within a company where there’s currently no app or automating that task before was simply not worth it.

If you could easily automate it then it’s probably already been automated. This AI is for the majority of tasks within processes which have never been automated because it was never viable or simple. Until now! We believe Trackable AI tasks fall into two categories:

• Let me show you (Navigational) – a preset way to navigate deep into an app or screen. This really helps when finding something or getting to the right screen is difficult. Perfect for starting out with navigational issues like “Where do I find this?” and “Show me where to go”.
• Do it for me (Completion) – a preset way to actually enter data, collect data and complete a task.

RPA stands for “Robotic Process Automation”. There’s lots of legacy vendors in that category. RPA was mostly only usable by large companies because it was very complex and very expensive. You needed armies of consultants and IT to make it work. Trackable AI + Computer AI is usable by anyone – including individuals running their own business. RPA bots tend to be expensive, brittle and flaky. They fail easily and silently. Which was never nice.

Middleware is a category of products that basically move data between one app and another app. They helps you integrate app without code. These days people are using middleware a lot less because it’s expensive to pay per-task and because AI now writes boilerplate code to integrate apps for you, making it much simpler and easier to handle custom integration. Examples of middleware apps are Microsoft Power Automate, Make, Tray, Workato and Zapier.

So – why is Trackable AI + Computer AI better than those?

There’s three basic approaches to automation:

• Documentation – Write down how a step is done and hope someone reads and actually follows the instructions.
• Attended Automation – Write a more detailed set of instructions to complete a task but empower a person to ask a system to do it. That’s what Trackable AI + Computer AI is. It gives you confidence and control because you can see it happening.
• Unattended Automation – It’s a black box you can’t see. You configure and create the automation and then just leave it to run. This works for a small set of invisible workflows, but it’s not ideal for most workflows in a business. It’s hard to configure, flaky and costly. Middleware and RPA fall into this category.

Computer AI + Trackable AI is Attended Automation. Here’s why it’s safer and better:

• You’re watching. It’s not a black box. Which means you’re still in control. It’s like still having a driver at the wheel of a car, even though the driver may not have to do most of the work. The driver is still essential.
• You’re accountable. If there’s a problem – what happens? With unattended automation through middleware and RPA platforms, many automations break due to small changes or even worse – fail silently. Since they were run by a bot nobody is accountable.
• It’s way simpler. Unlike RPA and middleware platforms where you’re writing code, hoping it works and then writing more and more code to patch problems – this is far simpler for business users. Simpler means better because it’s safer and not easy to break.
• It handles change. Because it’s powered by AI – it auto-handles most changes.
• It focuses on empowering people. Middleware and RPA are specifically aimed at replacing people for mundane tasks. We believe people are capable of adding a lot of value if they let AI handle mundane, admin and busywork tasks. In the end, you can also do more with less people but the quality of outcome, making a task mistake-free, standardized and training-optional for people is real empowerment.
• You start in minutes. Since instructions are in plain English – anyone in any role can use it without needing to be technical or have any specific expertise, unlike RPA and middleware.

You need an LLM vendor and you need Tallyfy.

The LLM vendor is your choice. Examples are Anthropic, OpenAI, Skyvern or even an open source LLM. Their pricing is separate and Tallyfy is not responsible for that. The cost of tokens can be unpredictable but that’s improving rapidly.

Tallyfy’s pricing has two elements.

• A per-person subscription that’s on our pricing page. The minimum’s just one full member.
• A per-minute cost for every minute Trackable AI runs on a computer.

We’ve made the pricing very simple and usage-based so that it’s accessible to everyone from individual freelancers all the way up to Fortune 1000 companies.

The ROI for Trackable AI is simple and clear – it completes a task mistake-free and quickly for a person, saving them time so that they can add value in other ways. Every minute it runs is a minute of reduced stress and consistent outcomes. We believe Trackable AI works best when it augments people and remains supervised by people so that people are still accountable for outcomes. The quality of an outcome is often more important than how fast you do it – especially if you’re new to a role or new in a company.