I would go even as far as to say that agency isn't desirable in most use cases. If only for the simple reason that when it comes to customer-facing applications, the biggest requirements for enterprises (mostly for compliance reasons) are reliability and control.
So, we’re essentially saying that entities will still need consultants who can code to build customized agents—unless their in-house IT teams have the skills to work with and code agents using GenAI. The idea of the slow demise of consulting companies seems premature. Consulting companies will likely adapt and thrive by acting as specialized architects for these advanced systems, bridging the gap between enterprise needs and the complexity of GenAI systems.
These agents might evolve to behave more like APIs interacting with each other in real-time, but they will be far more expensive than traditional APIs. This is because each call may involve not only a target system but also the computational overhead of LLMs. The combination of increased complexity and LLM involvement inherently raises costs. With this complexity, the chances of failure multiply, as many organizations struggle to make real-time APIs work reliably without robust middleware. Adding GenAI agents to the stack will only complicate this architecture further.
We should also prepare for agents failing to work or acting unexpectedly due to issues like LLM hallucinations or poor integration. Many entities will face challenges ensuring agents behave predictably, especially when LLMs generate incorrect or misleading responses that break workflows or lead to suboptimal outcomes.
Believe it or not, the big winners of the genAI boom are in fact the consultants:
“Accenture’s annualized GenAI bookings, from its Q3 stats, are $3.6 billion. To put this number in context, OpenAI’s annualized revenue is $3.4 billion. Accenture isn’t alone, either. Boston Consulting group, which made $12.3 billion in 2023, is projecting 20% of its 2024 revenue, and 40% of its 2026 revenue, to come from AI integration projects.”
I saw these statistics, but I also saw a post about consulting needs going down as people use LLMs to code and analyze data, etc. It looks like the new silver bullet(LLM), at least for the foreseeable future, would still need consultant help to build our use cases.
Great article! Thanks so much for touching on this topic. Refreshing.
There is not a lot that I would outsource to an autonomous agent that merely runs on LLM technology.
This is such an important piece.
We don't yet have good frameworks to design cognitive workflows that use nondeterministic software.
This year we're going to start to see some of these frameworks emerge and dominate over others.
The other key insight is that you don't need FULL agency to create business value.
Thanks for your response, David!
I would go even as far as to say that agency isn't desirable in most use cases. If only for the simple reason that when it comes to customer-facing applications, the biggest requirements for enterprises (mostly for compliance reasons) are reliability and control.
You're giving me ideas for a new post now. I'll write it up piggybacking off of this:D
Finding appropriate levels of agency for enterprise AI applications.
Can't wait to read it!
I may have misunderstood your post:
So, we’re essentially saying that entities will still need consultants who can code to build customized agents—unless their in-house IT teams have the skills to work with and code agents using GenAI. The idea of the slow demise of consulting companies seems premature. Consulting companies will likely adapt and thrive by acting as specialized architects for these advanced systems, bridging the gap between enterprise needs and the complexity of GenAI systems.
These agents might evolve to behave more like APIs interacting with each other in real-time, but they will be far more expensive than traditional APIs. This is because each call may involve not only a target system but also the computational overhead of LLMs. The combination of increased complexity and LLM involvement inherently raises costs. With this complexity, the chances of failure multiply, as many organizations struggle to make real-time APIs work reliably without robust middleware. Adding GenAI agents to the stack will only complicate this architecture further.
We should also prepare for agents failing to work or acting unexpectedly due to issues like LLM hallucinations or poor integration. Many entities will face challenges ensuring agents behave predictably, especially when LLMs generate incorrect or misleading responses that break workflows or lead to suboptimal outcomes.
Believe it or not, the big winners of the genAI boom are in fact the consultants:
“Accenture’s annualized GenAI bookings, from its Q3 stats, are $3.6 billion. To put this number in context, OpenAI’s annualized revenue is $3.4 billion. Accenture isn’t alone, either. Boston Consulting group, which made $12.3 billion in 2023, is projecting 20% of its 2024 revenue, and 40% of its 2026 revenue, to come from AI integration projects.”
Source: https://sherwood.news/business/generative-ai-consulting-war-block-trading-a24-creative-economy/
I saw these statistics, but I also saw a post about consulting needs going down as people use LLMs to code and analyze data, etc. It looks like the new silver bullet(LLM), at least for the foreseeable future, would still need consultant help to build our use cases.
Is there an A.I. Agent that can make the Gold and Bitcoin markets have 'number go up' indefinitely?
Everything that goes up must come down.