8 min read
How to Leverage GPT Functions in Your Products
A guide to GPT function calling and 5 useful use cases
Co-Founder/CEO
For an AI solution to be truly useful to a business it needs to be both reliable and affordable.
November 20th 2023Agent frameworks like AutoGPT and BabyAGI are what inspired AgentHub. To clarify, we're not knocking the autonomous agent approach. We simply chose to tackle the problem of automating work in a different way.
We've found that fully autonomous agent approach (at least in the short term) is too unreliable and expensive. This is especially true if you're trying to use these agents within a business or to address an uncommon usecase that requires specific context/integrations. We have taken a different approach to automation but one that could potentially be used in collaboration with these fully autonomous frameworks in the future!
Reliability is one of the key differentiators between a fun demo and a valuable product. Especially for business use cases.
The tasks we've seen businesses try to automate are almost always critical for day-to-day operations. If a workflow wasn't truly necessary they would opt to avoid it before spending time and money canvasing different AI solutions. There is very little room for error in these cases.
AI solutions need to not only match the cost of a human employee but be significantly cheaper.
If all other variables are constant and the prices are comparable, it's often easier to hire a human for the job. Humans are more versatile. They can be used to fill gaps in other departments, they add to the company culture, they can help out with physical tasks, etc. AI solutions need to be so much cheaper that the decision is attractive enough to endure any initial setup friction.
We're aware this point can feel a bit counter intuitive. We're an AI company advocating for using less AI but this opinion is driven by our thousands of hours building in the space. We only use AI when it is absolutely necessary in a workflow.
An average AgentHub automation is around 10% AI and 90% infrastructure. By infrastructure I mean the loading, manipulation and writing of data around these AI nodes. By relying on standard software for the infrastructure we can ensure that the critical AI nodes are at least getting the right context to perform their tasks and we can check/correct their outputs once they've completed.
One exciting opportunity we see on the horizon is for autonomous agents to use AgentHub automations as tools to complete tasks. AgentHub provides the framework for defining tasks that have a clear beginning and end. When to use these automations, how often and with what inputs is where autonomous agents come in. We imagine AgentHub automations being tools in an agents tool kit so it can take action reliably and solve more complex business problems.
Our automation building approach lets users explicitely define the exact order of operations they want performed for a task so there is no room for error. I like to imagine it like bowling with the child-proof guardrails up. You'll always be hitting the pins.
Here is some more content related to Gumloop, LLMs and automation. Enjoy!
8 min read
A guide to GPT function calling and 5 useful use cases
Co-Founder/CEO
3 min read
The journey of building AgentHub, from discovering AutoGPT to creating a platform for reliable and cost-effective AI automation.
Co-Founder/CEO
Automate your complex business processes without writing a single line of code. No credit card required.