I get excited when I see things like this, even if they're simple because I think I can build a business on top of it. However, in complex tasks and real-life cases, it's successful in very few instances. I can't trust its stability. This makes me feel like I've been deceived. I believe agents will be used for tasks that require very little intelligence and constant repetition. It's actually an assistant in situations like this demo. I want to use agents everywhere, but they're not successful in their outputs. GPT-4 is still being used at large scales. I don't know what the situation is with high-level usage of models that do reasoning like o1 through APIs, I haven't tried it. I tried Deepseek, and I encountered stability issues with Deepseek APIs. Besides, R1 doesn't have function calls.
I see no evidence that we're anywhere close to "fire and forget" AI that can be trusted to operate independently. But it feels like every business is centered on not only that being inevitable, but incredibly close if not already here.
Yet AI as a powerful tool utilized by skilled humans is already here, but we can't seem to shake free of this false promise.
I agree. The other day I went to an event. One of the speakers said he had a $2.7 billion exit. Everyone in the room believed it because some brilliant people and high-level authorities in the room believed in him and put him on that stage. Nobody thought otherwise because maybe this could be true. It's not logical to claim otherwise about what someone who might have $2.7 billion says (this really happened). Business, I don't know if there's a theory about this, but they think maybe what people who have received billion-dollar investments say might be true because saying things contrary to what they say doesn't gain anything.
Additionally, I don't want to be misunderstood - Agents have started making significant changes in the workforce right now, and we're even building an agent framework. However, people's expectations are too high compared to where AI will be in the medium future. This turns AI into hype. You will Remember what Sama said about Elon's AGI post.
This is pretty basic for a sales agent. Most of this flow has been available as sales enablement tech for > 10 years through Salesforce & Hubspot plugins.
Even Marketo could automatically place a call to a live warm agent with enriched information within 1 minute of a lead form being filled out on a website in 2010.
Operator did blow me away though. That is next level compared to Selenium and Puppeteer. While this sales agent feels very me-too.
I'm a consultant and my team has been building interactive "sales agents" using tech like Google Dialogflow, Amazon Lex, and Microsoft Bot Famework/LUIS since circa 2018-2019.
It does feel like we're reinventing the wheel with some of this stuff, except letting the LLM do things that used to be handled with deterministic logic in an orchestration layer.
Given the current capabilities of the tech, I'd still prefer to do things the old way and only inject the LLM at the point in the workflow where I need to actually generate content, like for the email body. Rather than giving the LLM tools and telling it to "figure things out" and execute end to end.
How much do those tools cost? And if they are gonna use OpenAI's tools anyway, adding a similarly capable sales agent at a fractional cost would be a win.
It's the same story that has happened with phone camera. It's became capable enough, at much lower price, since everyone already has a phone, that standalone camera is a rare sight nowadays.
y'all dont seem to understand. openai has built the most consequential brand for being frontier in AI. the demand for new AI tooling at all levels is through the roof. Hubspot and Marketo (lets call them the OGs) can add AI all they want, OpenAI will at least get a foot in that door, and yeah they will never be as focused on this customer profile as the OGs, but they will 1) be good enough for many 2) be first to market on capabilities that the OGs will not be 3) have context that the OGs don't (while also true that the OGs have context OpenAI will never have).
Even if this specific one flops, OpenAI has launched 2 agents in the first 2 months of this year. this is the 3rd. they're accelerating to a very exciting horizontal portfolio.
IMO it's the other way around – because OpenAI doesn't focus on any particular market, they'll be launching stuff that later will be easily reproduced by market leaders that have not only know-how, but they have the whole platform where the AI can be easily added.
What kind of big company will abandon Marketo because OpenAI launched an agent? OpenAI misses everything else that Marketo has, so they'll just wait a few months until Marketo adds the same feature
>IMO it's the other way around – because OpenAI doesn't focus on any particular market, they'll be launching stuff that later will be easily reproduced by market leaders that have not only know-how, but they have the whole platform where the AI can be easily added.
this is under appreciated. I remember people declaring Perplexity.ai doomed when ChatGPT search came out. Yet Perplexity is doing better than ever, with a service that lets you search with any major LLM, even DeepSeek R1.
Aidan McLaughlin, who now works for OpenAI, wrote a nice essay about this:
I am quite confident that this is basically a scam which won't work for at least 95% of businesses. Colin Fraser had a very nice demonstration that Deep Research can't keep track of NBA teams: https://xcancel.com/colin_fraser/status/1887020255781752949#...
So how exactly is a similar technology supposed to work with the highly specific details of an individual business? If the sales agent only has, say, 90% accuracy at reporting prices or product information, then it's unusable. I would guess it's actually much worse than 90%, even considering improvements in RAG benchmarks.
Maybe OpenAI figured it out and I will eat crow. But I suspect this presentation is marketed towards AI investors rather than sales executives. Investors are happy to pay for fun sci-fi stories about the future of AI, whereas salespeople actually need to sell something in the present.
Presumably the errors in the teams come from not having access to good data. The agent is reading websites and trying to make lists and sometimes going by memory - and, from the x thread, seems to often get things wrong.
What if the agent had access to good internal data? Instead of web scraping it queries a database or calls some API and does some filtering. Then it would be right roughly all of the time.
Have you ever wanted an AI slop plugin for your CRM? This seems really strange because it's obviously a good feature, if leads come in you definitely do want to be able to automatically get information on that prospect. Perhaps a call out to a linkedin API or some data broker who curates a reliable and detailed list of technical data on companies in your industry. There's clearly value in that. But what if instead of that useful reliable data, you put AI slop in there.
What Net Promoter Score impact will I have by having the first message I send to a prospective customer be AI slop?
I totally understand you want to AI slop loads of work, but the first interaction with a customer I would think would be a key time that you don't want to prioritize cost cutting?
Going by the most reliable thing, which is history, jobs will only become more abstract and more "fake busywork". As more and more of the real things gets automated, we moved to abstract things on top of the real ones. Managing the people / machines that do the real things for example. When you automate the second layer, the follow-up for me isn't "we'll go back to doing real things", but more like we'll create yet another layer on top and it'll suddenly be a very important new layer of work that "we must do".
It will be even more than that and this fits perfectly in the 2030 timeline which is exactly what I have been preparing for. [0]
These AI companies are experimenting with how much they can get away with mass job displacement as fast as possible; and they are very serious about it.
My company does AI Sales Lead Follow-up, and doing this ethically and effectively is a much more complicated problem than they suggest here.
Examples...
What time zone should we be following up in or offering times in
What salesperson should be assigned (aka whose calendar are we using)
Does the prospect ask a question in the form, do you trust an AI to answer that question?
Do you need to place a phone call?
Do you need to send a text (B2C)
How are you gonna do risk management (aka removing people who don't want more contact)
How does this interface with the existing systems
Are you going to let users change the prompts, how, what happens when they make a mistake
Many of these are messy sequential problems where solving each one reveals two more, and trial and error is the only way to get it right.
Which brings up the ultimate question, do you want someone to train their AI system on your sales leads?
This isn't as big of a breakthrough as it looks like, and I'm not sure who this is for.
There is already a heap of automation and tooling for inbound lead gen that ISRs use along with tools to easily research people/companies/org charts/etc.
Plug-ins for Salesforce also exist that do much of this once new leads are entered into the tenant (and they're also investing big dollars to AI-ify that flow).
Finally, inbound is a low-quality funnel compared to connecting through relationships or qualified leads from conferences and the like, so while, yes, this "could" augment a $45k/yr ISR role, inbound lead gen/lead qual is only a small part of what ISRs do.
Impressive demonstration, that’s clearly a huge help to sales folks. But I’ve already been ignoring a lot of the automated CRM emails that get sent out because they lack any sort of personal touch or interest, and I recognize it’s just an automated email to get me through their pipeline. Is this going to help with that sentiment?
Pretty much, kinda needs the next step there. Agent can do a bunch of API calls and report data out but what is missing is that it does not act based on those.
Meh, pretty basic. Most SaaS businesses have something like this in place already.
Closing inbound leads is relatively easy, since they've already shown active interest... The challenge I'm struggling with is (cold) lead generation: finding leads (and how to contact them) that match well with the service you're offering.
There are a lot of dubious scraping tools and B2B lead databases, but I feel like it should now be relatively easy to build a reliable web crawler & lead generator ... Does anyone know state of the art open source tools or services for this?
We are close to a point where bulk ai lead gen slop is going to be cheaper than smart targeting. I’d predict basically a slightly new / improved / arms race type incrementalism over the next 18 months. It will probably pay to stay on the bleeding edge, but it will be work.
I'm not in sales, but I imagine the customer attention is a limited resource in this scenario. If you create a fast, cheap and reliable way to generate leads, you'll saturate the market very fast. It's not really an edge if everyone uses it.
Curious why open source? At the end of the day, lead gen is mostly about data curation. You’re either paying for access to curated feeds or spending time/money building your own pipeline.
