Trust, Transparency, and Standards in Agent-to-Agent Markets

Julia Nimchinski:

Thank you. And with that, let’s give a warm welcome to Rena Jenna, Head of AI Research and Standards, Trust and Safety, Principal Evangelist and Editor at Google, and Allison Snow, Founder and Managing Director at Solicia Go-to-Market Advisory. And today’s topic is trust, transparency, and the standards in agent-to-agent markets.

Welcome, Allison, take it away.

Allison Snow:

Thank you so much. Good to see you, Scott and Randy, for a second there. Very quickly, just some context here. Welcome, Rena. It is so awesome to spend this time with you. As you know, everyone in the audience knows, HSE, does a lot of work in go-to-market.

They do, kind of, coaching there, a lot of this session is around go-to-market, and so when I was thinking about this session. I really just started to think about what’s on our minds today as go-to-market professionals, and what’s really current, because this is so current, this topic today.

And of course, we thank, longer sales cycles, bigger buying teams, just were always trying to sort of match the buyer’s behavior, or exceed it, exceed those expectations that keep evolving, and I just couldn’t stop thinking about how the oldie but goodie, if you will, is… is that, That, in fact, there is no buying experience outside of trust, right?

So, as we think about all of the tools we use, and all the things that we’re thinking about today, we’re just still trying to sell things and market things, and engage folks, and the trust infrastructure is really just what markets depend on.

And I was just reflecting that we see that in buyers increasingly asking more people for Evidence of the claims we make as vendors, buyers going to peer review sites. We respond to it by saying, hey, we’ll participate in your fantastic wave and magic quadrant and all of those things, and increasingly, as Scott talked about a little bit.

trying to appear in those LLMs and results as buyers increasingly do AI research.

And so we’re… we kind of get that infrastructure, but what’s so exciting about today is that all of that’s potentially about to change in the near or certainly not too distant future, as far as what we think about trust infrastructure, and I just couldn’t be… truly couldn’t be more pleased or excited.

I have people texting me saying, hey, I hope your conversation goes well today, I know you’re looking forward to it, to talk to you, Rena, just the perfect person to kind of ask about these changes and kind of help us understand in our own roles why all of this matters. So, welcome, and thank you.

Reena Jana:

Thank you so much, Allison and Julia, and it’s a real honor to be here with this group. All of you as CMOs, CEOs, senior executives, you’re all pioneers in this space, you’re all innovating, you’re facing challenges that no one else has faced before in the GTM arena, and you’re making decisions that no one else has made before.

And so, it’s exciting to be able to have this conversation with you, and sort of talk it through together on, you know, some trends that we’re seeing, and just Sharing some of the lessons learned that we’ve faced, too.

Allison Snow:

Excellent. It’s really so nice to be here with you, thank you, and on behalf of the community, thank you for your kind words about our innovation. I did actually put together some questions. I do typically try to make this conversational, but there were a lot of things that I just was afraid I’d forget.

So, I did want to hit on, again, the conversation we’re having about AI so much, and on this session. Previous sessions as well, and throughout this whole summit, is about productivity and being faster and doing more and faster workflows.

And we’ve seen some of it, but I think we’re forward-looking today, particularly you, thinking through the harder questions about trust and governance and privacy and what that trust infrastructure looks like. as buyers and sellers evolve, and Agentic evolves. So, just to start there.

Again, we’re thinking about agents participating in evaluating, procuring, negotiating, giving demos to each other. I don’t want to make this too pie in the sky, but what assumptions about trust are being tested first here?

Reena Jana:

Yeah, it’s a great place to start, and, I think Implied in your intro, you know, we’re really at the early stages of this technology. So early, in fact, I’m… I sit in a lot of expert groups and working groups on AI safety standards.

And just now, literally, like, this month, in the Organization for International Standardization, or ISO, We are just now working on international standard definitions for really basic terms. like, Agentic AI as a research and development discipline. Agentic AI system and AI Agent.

I know that all of your companies are already thinking about how do we deploy this?

How do we adjust for GTM strategies with Agentic AI, agentic AI systems, and AI agents, but we’re really at the beginning stages of, you know, thinking through best practices for trust and safety, thinking about just shared vocabularies around the world that are going to be important for interoperability in terms of, you know, safety policies, or even security policies, etc.

So, it’s only natural, since it’s so early, that questions related to trust are going to arise.

But I think to sort of put, all of your minds at rest, as you’re making decisions, as you’re leading teams in this space, I think it’s really important, we talk about this internally, too, and with customers, is that the world of Agentic AI is happening in the continuum of technology, it’s happening in the continuum of marketing strategies, as the previous talk focused on as well, towards the end, and what you were saying, and I think what is raising a lot of trust questions is the fact that AI is starting to evolve.

From simply answering, right? Answering customer queries, answering questions and needs and, etc, to acting, acting on behalf of us. And, While that is true. it renews a lot of existing trust questions, and I think that’s really important.

It’s easy to get, like, caught up in, like, thinking of all of these future sci-fi scenarios, but some of the eternal questions remain in terms of, obviously, privacy and security, content safety. impacts on, broader systems, digital commerce, etc.

But I think what’s really important to focus on now, it’s gonna sound like I’m going backwards in time, but I’m not really, are some of the specific risks that came up through, the sort of Gen AI revolution, that are still in play in terms of trust and safety with agents, so… really simply, Classic content risks, Agents could include or generate content that is harmful.

mistakenly offer profanity or a sort of tone that’s off. That could lead to brand safety risks, where agents’ generated content could, misalign with your brand’s tone. Editorial guidelines, your brand’s values. It could endorse a competitor by mistake, or an inappropriate product.

It might be using a sort of generic name for a type of product that you have, but is known, maybe popularly, associated with another brand. It could generate content that could result in reputational damage.

totally inadvertently, because it’s acting on its own with some level of autonomy, it could, you know, have those inaccurate outputs, hallucinations, etc. And then also, unintentionally, like, before, we used to worry about people posting content online or, or, creating content that, Might be… that might leak, sort of.

company secrets, or sensitive data, or, you know, system prompts, but agents can do that as well. So just thinking about the… how those risks are, magnified with autonomy. And then also, how bad actors. Bad actors are innovators, too. Say if somebody wanted to give a command to an agent to override safety protocols or behave in unintended ways.

So, I don’t mean to end on a negative note.

Allison Snow:

Bec.

Reena Jana:

But there are, you know, those risks remain evergreen. And then there are some other trust concerns that are unique to agents, that might be.

Allison Snow:

Yes.

Reena Jana:

of interest to this audience, you know, agents can… I’m sure you’ve been talking about this all morning, but processing tasks at, a speed and scale that we haven’t seen before? So think about your website traffic. Can I handle, you know, increased traffic?

Think of all the technical permissions, digital protocols, all of these things that are happening, with increased agentic, capabilities.

So, you know, again, I don’t want to make anybody paranoid, but really getting ready, like, balancing some of the existing risks, and then thinking about, the ones that are specific to agents, as you’re starting to consider how to deploy them.

Allison Snow:

Yeah, I think that, a few things. I… even though I was… I wrote down, bad actors are innovative, too. I think it’s a really interesting quote. That being said, I don’t think you made us paranoid or brought anybody down. In fact, I think if everybody had their volume on in this audience, we may have heard some sighs. of relief.

And the conversations I have with folks in GTM every day, peers, customers, friends, there’s this… constant feeling of being behind, and also, I think, just because of the headlines we’re surrounded by, also this constant suspicion that maybe all of the things that you just described in terms of risks have in fact been solved, and we just haven’t figured out how to solve them yet.

So I think there’s… Not paranoia, but really comfort, and you and your position saying, these are things we’re still figuring out and talking about at the highest levels of governance. For this item.

Reena Jana:

Yeah, I mean, there, they’re… I… I… wish I didn’t have to say this, but sort of evergreen… content and safety and security and privacy risks. I think it’s just the speed and scale of, of agents’ actions are something that, you know, exacerbate those risks, and are something to keep in mind.

I also just want to add, in terms of, sort of peace of mind in this era. In my opinion, I think it’s really, really important as senior leaders to think about, and also in your guidance to your customers and buyers, when they’re using agentic services that you might have on your platforms, is to Keep remembering.

the main purpose of agents, which is to take directions and complete actions on behalf of people, right? Via reasoning, which These models can do, meaning breaking things down into, different steps, planning towards achieving those goals, using tools, and a tool can be something like a search engine.

You know, it doesn’t necessarily mean something, Bigger than that. But, these agents are built… you’re building them, you’re deploying them, because you want them to act on behalf of a person, a buyer, your organization.

And I think, sometimes in the conversation around agents and autonomy, we… we lose sight of the fact that Agents are acting on behalf of people.

Allison Snow:

I think that’s important, I appreciate that. I wonder if you have thoughts yet, because it’s clear that this is all under discussion, again, in some of the groups to which you belong and participate in. and, and lead, in terms of disclosing when it’s an agent, right?

So… I think we’re all looking ahead and saying that it’ll be increasingly sophisticated, but even now, if it’s not kind of selling and negotiating, and rather a piece of content was written by AI, or an agent, you know, did do a… all of which to say, how much do you think people need to know right now?

If we’re being responsible actors, and we’re saying we want this to act on behalf of me as a person, is there a… A point at which you say, disclose, Or any guidance?

Reena Jana:

Yeah, that’s a big question. Clearly, there are a lot of disclosures out there, and also different technologies, even with GenAI, in terms of when you’re interacting with AI, when a piece of content was… generated by AI, etc. So, you know, this idea of clear notices, at the appropriate time. are important.

I’m just going to give you a preview of some research that my team has done. I have UX researchers on my team that did a survey in the U.S. Potentially, we might expand it on calibrating trust in Gen AI. In the U.S, and one of the questions we asked about was disclosure.

I think, yes, people want it, but they’re, they’re, they, they’re craving more. This was a survey of about 1,500 people in the U.S. across, you know, demographics that were, in parity with the U.S. Census.

And one of the findings was people wanted… yes, they wanted disclosure, but they also wanted guidance on how to use AI tools, or how to understand how it works. So, I think, you know, people are getting sophisticated. Like, they, yes, they want, they want disclosure, but they also want more than that, which is kind of interesting.

Allison Snow:

I mean?

Reena Jana:

That might be an area… again, this was a research, a research paper with the Advertising Council, and I’m happy to share it with you if you want to distribute it after After this, talk. But, but yeah, so disclosures, yes, but also more than that.

In terms of transparency, though, there are new issues that are coming up, with… with agentic experiences, as well. Obviously there are a lot of, Moments for users to consent, to give their consent.

Allison Snow:

So, in terms of…

Reena Jana:

privacy, or their permission to use a certain tool, etc. And this, is a real challenge in terms of, I would say goes beyond disclosure, but more into, sort of, these.

Allison Snow:

Right?

Reena Jana:

UX issues around, transparency, which I think is related to disclosure, that are… that are really tricky, you know? Like, how… how often do you want to Sort of engage with the user, the person, when the whole intent of the agent is that it acts on your behalf. Right.

Do you overburden people with disclosures or, you know, UX that asks for their consent as the agent is creating actions? There are a whole number of, of, sort of transparency and disclosure-related.

Allison Snow:

Okay.

Reena Jana:

Concerns that have now been raised.

Allison Snow:

Yeah, and I think so, so important for us as a GTM community, again, one thing that I think makes us always cringe a little bit is any additional friction to the buying process. So anytime someone has to make a decision, I mean, it could be another field in a form, and our inclination is no. You know, make the conversion point, whatever it is,

Reena Jana:

Exactly, yeah.

Allison Snow:

much smoother, so…

Reena Jana:

Yeah.

Allison Snow:

Those would be important questions.

Reena Jana:

Yeah, I think that’s… that’s really important, as you, as you build these, agentic experiences, or, you know, get onto certain platforms that allow for that. You know, how seamless is that customer journey? Right.

Allison Snow:

Right.

Fewer decisions, the better, and as you point out, and I think we all know, someone who is deeply, deeply educated on this stuff might say, I know exactly what this means, and I’m willing to consent, and then for some percentage of the population, of your buyer population, there’s just a, you know, more than an additional form field, there’s a decision of… I don’t fully know what I’m giving away here.

So that’ll be… we might think about investing in some education around that as brand ambassadors. Whole different story. You mentioned that in that survey, folks said, you know, disclosure and maybe more.

I’d just like to take a guess at what more could look like, and then you could say, you’re totally wrong, and let folks know for real from the real expert. I imagine that some of that is, hey, I not only want to know it’s an agent, but also who authorized it, and what are they optimized to do?

Is that what Moore looks like for them, or… tell us the real deal.

Reena Jana:

Not that granular yet, but I think that that is a concern that, is being addressed with different types of approaches to, either sort of technical protocols, between agents, or, with user interfaces. I… I will say, in that specific research, it really was more, sort of, instructions.

on how to use the technology, which I think is really great, because I think people…

Allison Snow:

And again.

Reena Jana:

And as I emphasized earlier, agents are acting on behalf of people, and so the more clear a user is of an agent, in what they want the agent to achieve, the better results they will get, and the more they’ll be able to rely on that agent.

I hope that answers your question, but I can also talk a little bit more about, sort of transparency around agent identity and some trends that we’re seeing that might be of interest to this group.

Allison Snow:

I think that would be great, and I think what folks are really, you know, you’re hitting it on the head. I mean, people are kind of saying, what is everyone else doing?

Reena Jana:

Yeah.

Allison Snow:

how am I not behind? You know, I’m just trying to put some… some words around what I think people on this forum are thinking, and you know, what is the best practice to which I don’t have access? I would love that, I would definitely ask that of you.

Reena Jana:

Yeah, yeah, in terms of the, sort of, agent, identity and, like, learning more about an agent or what’s out there in terms of building trust in, agents that are maybe interacting with each other.

You… in the audience, you might be familiar with, and maybe you’ve talked about it earlier today, but, protocols like, A2A, agent-to-agent, I think that might even be the name of the, this, this, session, but, there is a protocol, called Agent 2, the number 2 agent that, that Google developed, and it’s out there now, where agents actually, carry something called agent cards.

They’re almost like… Calling cards or license plates, just to talk about really old technology, but inspired.

Allison Snow:

Bye.

Reena Jana:

By those, that will contain things like The agent’s name, who provided it.

Allison Snow:

and…

Reena Jana:

what it’s capable of, and other information. That’s for agents to communicate with each other, but already there are, sort of, spurly systems in place to offer that type of transparency and information, again, agent to agent, to, sort of.

Build that trust behind the scenes, which… I would, you know, recommend maybe seeing if some of the platforms you are, if they’re using that A2A protocol, then you will know, you know, oh, those agents have those agent cards that might give a little bit of trust. In that type of platform.

Allison Snow:

Yeah.

Reena Jana:

Yeah, and, you know, the nice thing is there are a lot of these sort of behind-the-scenes, approaches that, again, you might not see as a GTM leader, but, and the, the, sort of engineers and behind the scenes, like, choosing platforms that have, those types of protocols built in, or different tools, you know, we’re seeing things like, I can only speak for Google, but, you know, we have an enterprise agent platform that lets, developers who use that platform actually assign unique identities to different agents that they are building.

So, there’s that sort of accountability. You can sort of trace what they’re doing. So again, these sort of behind-the-scenes trends.

Allison Snow:

That’s inspirational. I think what folks might wonder, or what you’ve inspired me to wonder, and I think they will, is if we are… folks who are planning to put agents in customer-facing workflows at all.

Before we even think about, transparency, and this is an agent, which I don’t think makes anyone excited right now, Is there anything internal governance-wise? I mean, you just named one to sort of, the traceability, right, of a responsible person that said, act on my behalf, and having them linked to the agent.

And it made me wonder if there’s any other governance standards or guidelines that, that you may have seen in these meetings, or… Oh, yeah, yeah.

Reena Jana:

Yeah, this might… I don’t know if this seems boring, but I hope not. I hope it gives a little bit of, Sort of comfort and, and trust, I think the AI governance and agent governance and internal governance, it’s still, it’s becoming more standardized. Like I say, there are frameworks that still apply to building agentic systems, like, NISTs.

AI, Risk management framework is one. Another on an international level for, for those of you on the call that have really broad international markets, ISO’s standard 42001 is an AI risk management system standard that companies, organizations can actually certify against.

to show that they have responsible AI, risk management in place, so it’s a very helpful way of either showing if you’re building an AI system. you can get certified against this coming… AI governance is getting professionalized, basically.

Allison Snow:

Yeah.

Reena Jana:

So, you know, your company or your organization can get certified against it, and you can also check if you are using, a… you know, a product, an agentic product, that has been.

Allison Snow:

Certainly.

Reena Jana:

against this, so…

Allison Snow:

Yeah.

Reena Jana:

We’re going a little bit beyond, you know, sort of Aspirational frameworks to, sort of, the professionalization of,

Allison Snow:

Yeah.

Reena Jana:

and governing.

Allison Snow:

recognized organizations.

Reena Jana:

Yes, yeah, so that’s.

Allison Snow:

Which is fantastic, yeah.

Reena Jana:

And that’s been in play for a while. While it might not be agentic-specific, it is still Applicable, and might, might, you know, put, put, your team’s minds and your own mind at ease,

Allison Snow:

Yeah.

Reena Jana:

Knowing that those frameworks are in place, and, like, really professional certification, too.

Allison Snow:

Yeah, and I, you know, I always have my marketer hat on, but I have a buyer hat as well, right?

Reena Jana:

Yeah, exactly. Exactly, yeah.

Allison Snow:

Yeah, so it’s important, I think to know that even buyers will have a place, because right now, again, in our current lives, we may say we want to be on G2, Trust Radius, and those are sources of trust, and certainly for things like technical infrastructure, there are ISO being one of them.

standards in the world, but we don’t have a way, necessarily, I think, down to objectively give buyers comfort or get comfort as buyers like that, so I think that’s… I didn’t realize all that was happening, so… Maybe some of our viewers didn’t need it, that’s great. Do you have an opinion?

I think… Some of us on the call might say, hey, I’m being kind of slowed down and challenged. By the folks in my organization who care about privacy and trust, and there’s… I think this… Evergreen, if you will, you know, things that never change.

Forces between innovation, doing the next thing, being ahead, and creating comfort in your own organization with the folks who are ultimately responsible for being in compliance with many, many things.

Do you… do you see that… that… that force, and Google, if you could talk about it, or in unknown companies that you might have access to, and information about? And what does that usually look like?

Reena Jana:

Well, I mean, you know. I think I said this earlier, like, AI agents, they’re a part of a continuum of.

Allison Snow:

Yeah.

Reena Jana:

artificial intelligence, of online experiences, etc. So I think, you know, once you start Thinking in this way, you can start applying frameworks that have worked in the past. So I think I’ve been implying that, as we’ve been chatting now.

It’s like, okay, move fast, their agents are coming, they’re here, they’re moving ahead, but how do you sort of apply enduring frameworks? It’s people using, these, these systems. You are… going to market with products for people. There are the same concerns, again, to echo the previous, talk as well.

So, you know, bringing… bringing that to the table in terms of even from the trust and safety side of things, it’s sort of like, okay, we don’t have to start from the beginning, but we want to make sure that while we’re, thinking of what’s new, we’re always thinking about, like, what can we still apply?

I think that’s how, at Google, we’ve been able to, sort of evolve quickly, where responsible AI sits in the Trust and Safety Organization, which actually has been around for almost, you know, has almost 20 years of experience of fighting online abuse. And, you know, again, there’s no sort of finish line.

New technology is bringing new risks, but Some of the… those, essential risks are… are still there, and we can apply what we’ve learned to that. I think the good thing, though, just again, to, maybe offer a little bit of, peace of mind in terms of, innovation.

Like, safety innovation is innovation, too, and the cool thing about AI and the models is you can bring that in very early on the model level, sort of bake in policies and, train models and fine-tune them and, Use reinforcement learning with, with, you know, human supervision, et cetera, to… to abide to safety principles, right?

And automate that sort of safety, from the beginning. And so, that’s actually kind of exciting. I think some people, maybe in, there’s a cliche that, like, oh, safety, you know, sort of slows things down. It’s actually, actually, you know, safety is an innovation too, and you can bake it in a.

Allison Snow:

Right.

Reena Jana:

Earlier in the… in the product development lifecycle, or the, content platform development lifecycle, or marketing podcast.

Allison Snow:

Totally.

Reena Jana:

Lifecycle on the model level, which is exciting. Like, it’s a story that doesn’t get told as much.

Allison Snow:

Yeah, no, I think the idea that innovation and safety are not natural enemies is really an important takeaway, because I think It’s… it’s worth reminding, right, that we’re all in this together.

Reena Jana:

Yeah.

Allison Snow:

And we’re all accountable for different things, so that’s… that’s excellent. I know that we’re 2 minutes away from the end, so I want to make sure, Irina, if there was anything you wanted to make sure you told this audience before you Before you, you, you stopped gracing us with your presence, what, what would it be?

Reena Jana:

Yeah, I think, what we’re really seeing is that there is just such a need for domain experts to be involved in.

Allison Snow:

Hmm.

Reena Jana:

the whole AI trust and safety standards space. So, as I was saying earlier, everyone here today, is an expert in GTM. You are all innovators and pioneers in how Agentech AI will play a role in the go-to-market world. Bring your point of view to the table.

thought leadership in this space, thinking about, like, how to become trusted testers of these Agentic AI systems. What are the challenges you’re facing? What is the point of view from the GTM community? I think it’s really, really important.

So I would encourage, everybody here to you know, be thought leaders in the space that you know well, and, really surface.

That’s why I’m just so honored to be here and to have this conversation with you to understand, you know, like, what are some of the needs of the GTM community, as, we and others continue to develop, the Agentic AI systems together, again, with.

Allison Snow:

Yeah, it’s a great call for us to participate.

Reena Jana:

Yeah.

Allison Snow:

Thank you so much.

Reena Jana:

Yeah.

Allison Snow:

Thank you.

Reena Jana:

My pleasure.

Julia Nimchinski:

Fantastic conversation. Thank you so much again, Allison and Brina. And for all the folks watching, what’s the best way to support you, really briefly?

Reena Jana:

Oh, yeah, see, I think, again, just sharing your wisdom in terms of, GTM concerns, become active, and there are a lot of multi-stakeholder, organizations in AI development. I’m happy to share out some of the resources with you if you’re doing a follow-up email to this conversation. And we can take it from there, but thank you so much.

Julia Nimchinski:

Awesome. Thank you, Rena. Allison?

Allison Snow:

Sure, I write a sub-stack called The Alignment Dividend. If you are interested in why GTM Harmony wins, go ahead and give that a follow. I’d love to share my thoughts with you. And please feel free to forward me all of those things that you’re writing and authoring, so I can support you as well.

Julia Nimchinski:

Thank you so much, Allison.

Allison Snow:

Yeah, thank you.