Select Page
artificial intelligence 4389372 1280
In-House And AI: You Can’t Demand What You Don’t Understand 3

It’s a mystery. Most in-house counsel want their outside firms to aggressively use AI, and most expect their bills to be reduced. Yet, they rarely demand these things from their outside firms. It’s a counterintuitive gap. What gives?

A recent Axiom study sheds some light. Put simply, a lot of in-house counsel don’t have any more of a clue how to use AI effectively than do their outside firms. If the people paying the bills don’t understand what AI can realistically accomplish, how can they know what to demand? How can they hold the people sending the bills accountable?

The Study and Its Source

Now we do need to start with a little healthy skepticism. Axiom is an alternative legal service provider (ALSP). Its primary mission is to supply in-house legal departments with legal talent for specific and often temporary needs. So it’s in Axiom’s interest if in-house legal needs help with AI. That does not necessarily mean the survey results are wrong. I have no evidence of that. But it’s always good to keep in mind who is paying for work like this.

Axiom sponsored the study and commissioned InsightDynamo, which describes itself on its website as a healthcare research firm (Hmm …). InsightDynamo surveyed 528 in-house legal professionals online at companies with annual revenues of at least $100 million. The survey was done between March 13, 2026, and March 25, 2026.

A Lack Of Metrics, Skills, And Use

The results paint a bleak picture of in-house legal’s understanding and effective use of AI. Of those surveyed, 83% don’t know how to measure AI’s ROI or assess how their department’s AI spending is working out. That means fewer than 20% have established metrics to assess the effectiveness of what they’ve spent. 

Only 7% are optimizing and measuring AI across their organization. Think about that: they’re spending lots of money on AI but have no clue if it’s producing meaningful returns.

Most respondents assessed their AI skills as mainly intermediate. Only 10% believe that they are advanced. There’s a reason for that: only 25% send their personnel out for external training. Axiom notes that departments are increasing AI spending “with a workforce that has had at most one working day of training.” Jeez.

Only 17% of those surveyed report that a majority of their workers use AI tools daily. According to the report, only one in four use AI monthly. As Axiom puts it, “the skills gap is not access or adoption; it is training and regular experience using the tools.” Only 40% have implemented AI planning and governance practices. Another 40% say they plan to, which means they haven’t yet. (Or as they say where I grew up, they’re fixin’ to. Right.)

The barriers that the respondents identify to determining AI value are also instructive. They include:

– a belief that the workflow changes would be too disruptive

– a feeling that team members will resist change

– the view that skills are lacking

– a sense that ROI is just too hard to measure

Sound familiar? It’s the same list you’d probably get from a room full of law firm managing partners. 

Nonetheless, 95% magically expect outside counsel to use AI (and use it effectively), and 92% expect lower rates when firms do use AI, an expectation that is largely unsatisfied. But another survey shows that 80% of in-house don’t require or even encourage their firms to use AI.

Legal AI? What Legal AI?

And what AI tools are being used? 67% use general-purpose or frontier models, not legal-specific models. Tools like ChatGPT, Copilot, and Gemini. The same tools most of us use for nonlegal work. As the report puts it, “Most legal teams are running AI the same way a marketing department would.” That suggests most legal departments have no idea what they’re doing with AI. 

Selection Process? What Selection Process?

And the selection process for AI tools? In a word: abysmal. Only 1 in 10 teams conduct rigorous evaluation before determining what product to buy. Fewer than 1 in 4 run structured pilots with any sort of defined goals, timelines, or measurement. 

If you don’t know what you want a tool to accomplish before you buy it, it’s pretty hard to determine whether it accomplished anything afterward. It’s no wonder the ROI is elusive.

Guidance? What Guidance?

And in-house teams are desperate for guidance. Almost all, 98%, say they wish they had outside guidance on tool selections, and 96% say that they would use an ALSP for adoption or implementation advice, not a law firm. This is probably because ALSPs generally have an economic incentive to use AI and are therefore perceived as further along the learning curve. It also reflects a lack of trust in law firms. (Axiom is an ALSP.)

What Does It All Mean?

It’s not surprising that legal departments aren’t seeing their bills reduced. In-house legal departments don’t really know how to use AI effectively. The numbers clearly show a real lack of knowledge and understanding among in-house legal teams when it comes to AI. They can’t figure out the return on investment, which means they can’t identify where AI would be valuable or how to get there.

If they can’t figure out the value, how in the hell can they tell outside counsel how to use the tools in ways that reduce their bills? How can they have a meaningful conversation with outside counsel about where AI should be used to reduce fees without sacrificing quality?

Moreover, outside counsel clearly won’t use AI to reduce their bills until clients demand it. First of all, most don’t have any more of a clue about its use than in-house. More importantly, those firms that bill by the hour (as do many of the outside firms that work for these companies) have no economic incentive to use tools and technologies that they fear will ultimately reduce their revenue. As Richard Susskind famously noted, it’s hard to tell a room full of millionaires their business model is all wrong. It’s even harder to convince them.

In-house counsel, on the other hand, have the economic incentive to use AI. But they aren’t using it as effectively as they should. They don’t know which product to use, nor do they know how to conduct a robust evaluation to decide what to use. And just like their outside brethren, they view AI use as disruptive and resist change. 

So they’re not much different from the outside firms they now expect to use AI despite those firms’ disincentive to do so. Perhaps that expectation is not unreasonable, since outside firms are in the business of serving their clients. So it would be in their clients’ best interests for them to use AI. That may be true for some firms, but that’s simply not reality for most.

The Blind Leading The Blind

So here we sit in 2026. You have law firms reluctant to use AI and fearful of what it could mean for their economics. You have in-house legal departments that don’t understand AI’s value or how to use it effectively. That creates a stubborn gap between what in-house counsel want and what outside firms are doing. 

But unless and until in-house legal departments better understand how to use AI and what can be realistically demanded of outside firms, that gap will remain. You can’t demand what you don’t understand.

Want your outside firms to use AI effectively? Learn how to use it in-house and then demand that they do the same. 


Stephen Embry is a lawyer, speaker, blogger, and writer. He publishes TechLaw Crossroads, a blog devoted to the examination of the tension between technology, the law, and the practice of law. 

The post In-House And AI: You Can’t Demand What You Don’t Understand appeared first on Above the Law.

artificial intelligence 4389372 1280

It’s a mystery. Most in-house counsel want their outside firms to aggressively use AI, and most expect their bills to be reduced. Yet, they rarely demand these things from their outside firms. It’s a counterintuitive gap. What gives?

A recent Axiom study sheds some light. Put simply, a lot of in-house counsel don’t have any more of a clue how to use AI effectively than do their outside firms. If the people paying the bills don’t understand what AI can realistically accomplish, how can they know what to demand? How can they hold the people sending the bills accountable?

The Study and Its Source

Now we do need to start with a little healthy skepticism. Axiom is an alternative legal service provider (ALSP). Its primary mission is to supply in-house legal departments with legal talent for specific and often temporary needs. So it’s in Axiom’s interest if in-house legal needs help with AI. That does not necessarily mean the survey results are wrong. I have no evidence of that. But it’s always good to keep in mind who is paying for work like this.

Axiom sponsored the study and commissioned InsightDynamo, which describes itself on its website as a healthcare research firm (Hmm …). InsightDynamo surveyed 528 in-house legal professionals online at companies with annual revenues of at least $100 million. The survey was done between March 13, 2026, and March 25, 2026.

A Lack Of Metrics, Skills, And Use

The results paint a bleak picture of in-house legal’s understanding and effective use of AI. Of those surveyed, 83% don’t know how to measure AI’s ROI or assess how their department’s AI spending is working out. That means fewer than 20% have established metrics to assess the effectiveness of what they’ve spent. 

Only 7% are optimizing and measuring AI across their organization. Think about that: they’re spending lots of money on AI but have no clue if it’s producing meaningful returns.

Most respondents assessed their AI skills as mainly intermediate. Only 10% believe that they are advanced. There’s a reason for that: only 25% send their personnel out for external training. Axiom notes that departments are increasing AI spending “with a workforce that has had at most one working day of training.” Jeez.

Only 17% of those surveyed report that a majority of their workers use AI tools daily. According to the report, only one in four use AI monthly. As Axiom puts it, “the skills gap is not access or adoption; it is training and regular experience using the tools.” Only 40% have implemented AI planning and governance practices. Another 40% say they plan to, which means they haven’t yet. (Or as they say where I grew up, they’re fixin’ to. Right.)

The barriers that the respondents identify to determining AI value are also instructive. They include:

– a belief that the workflow changes would be too disruptive

– a feeling that team members will resist change

– the view that skills are lacking

– a sense that ROI is just too hard to measure

Sound familiar? It’s the same list you’d probably get from a room full of law firm managing partners. 

Nonetheless, 95% magically expect outside counsel to use AI (and use it effectively), and 92% expect lower rates when firms do use AI, an expectation that is largely unsatisfied. But another survey shows that 80% of in-house don’t require or even encourage their firms to use AI.

Legal AI? What Legal AI?

And what AI tools are being used? 67% use general-purpose or frontier models, not legal-specific models. Tools like ChatGPT, Copilot, and Gemini. The same tools most of us use for nonlegal work. As the report puts it, “Most legal teams are running AI the same way a marketing department would.” That suggests most legal departments have no idea what they’re doing with AI. 

Selection Process? What Selection Process?

And the selection process for AI tools? In a word: abysmal. Only 1 in 10 teams conduct rigorous evaluation before determining what product to buy. Fewer than 1 in 4 run structured pilots with any sort of defined goals, timelines, or measurement. 

If you don’t know what you want a tool to accomplish before you buy it, it’s pretty hard to determine whether it accomplished anything afterward. It’s no wonder the ROI is elusive.

Guidance? What Guidance?

And in-house teams are desperate for guidance. Almost all, 98%, say they wish they had outside guidance on tool selections, and 96% say that they would use an ALSP for adoption or implementation advice, not a law firm. This is probably because ALSPs generally have an economic incentive to use AI and are therefore perceived as further along the learning curve. It also reflects a lack of trust in law firms. (Axiom is an ALSP.)

What Does It All Mean?

It’s not surprising that legal departments aren’t seeing their bills reduced. In-house legal departments don’t really know how to use AI effectively. The numbers clearly show a real lack of knowledge and understanding among in-house legal teams when it comes to AI. They can’t figure out the return on investment, which means they can’t identify where AI would be valuable or how to get there.

If they can’t figure out the value, how in the hell can they tell outside counsel how to use the tools in ways that reduce their bills? How can they have a meaningful conversation with outside counsel about where AI should be used to reduce fees without sacrificing quality?

Moreover, outside counsel clearly won’t use AI to reduce their bills until clients demand it. First of all, most don’t have any more of a clue about its use than in-house. More importantly, those firms that bill by the hour (as do many of the outside firms that work for these companies) have no economic incentive to use tools and technologies that they fear will ultimately reduce their revenue. As Richard Susskind famously noted, it’s hard to tell a room full of millionaires their business model is all wrong. It’s even harder to convince them.

In-house counsel, on the other hand, have the economic incentive to use AI. But they aren’t using it as effectively as they should. They don’t know which product to use, nor do they know how to conduct a robust evaluation to decide what to use. And just like their outside brethren, they view AI use as disruptive and resist change. 

So they’re not much different from the outside firms they now expect to use AI despite those firms’ disincentive to do so. Perhaps that expectation is not unreasonable, since outside firms are in the business of serving their clients. So it would be in their clients’ best interests for them to use AI. That may be true for some firms, but that’s simply not reality for most.

The Blind Leading The Blind

So here we sit in 2026. You have law firms reluctant to use AI and fearful of what it could mean for their economics. You have in-house legal departments that don’t understand AI’s value or how to use it effectively. That creates a stubborn gap between what in-house counsel want and what outside firms are doing. 

But unless and until in-house legal departments better understand how to use AI and what can be realistically demanded of outside firms, that gap will remain. You can’t demand what you don’t understand.

Want your outside firms to use AI effectively? Learn how to use it in-house and then demand that they do the same. 


Stephen Embry is a lawyer, speaker, blogger, and writer. He publishes TechLaw Crossroads, a blog devoted to the examination of the tension between technology, the law, and the practice of law.