I interviewed the CEOs of the hottest AI sales agent startups.

Here's what AI sales agents can (and can’t) do for you.

Welcome back.

As a reminder, I’m Chris Bakke.

Today I’m very excited to feature an incredibly well-researched guest post from Lauren Vriens. I’ll let Lauren take it away.

-

Hello, I’m Lauren Vriens - AI obsessor, startup advisor, and operator once famous for scaling a moped startup to $50M in a city where no one knows how to drive.

In today's Tech News, But Funny guest post, I'm diving deep into AI sales agents - the hype versus the truth. If you like it, let us know. If you need help turning AI buzz into real business growth, hit me up at vriens.lauren@gmail.com or www.linkedin.com/in/laurenvriens.

I. Introduction

You may be a founder trying to spin up a sales team and burning through your precious runway. After 3 cold brews and 5 mg of Zyn, you may have had the thought: "if AI is so darn great, why can’t it do sales for me?”

Y Combinator and countless VCs are betting big on AI for outbound sales. Makes sense. We’re entering an era where ads have diminishing returns, copy cats launch in a weekend, and every market is more crowded than Trader Joe's on a Sunday.

But let's get serious: how realistic is it that AI agents today can do sales?

In this deep dive, I’ll explore the world of AI sales agents - the good, the bad, and the downright illegal.

I’ll cover:

• The current landscape of AI for outbound sales

• What AI sales agents can and can't do

• Profiles of 3 darling AI sales agent startups

• How to evaluate potential solutions

• What the future holds for AI sales

The TL;DR on AI Sales Agents Today:

1. AI sales agents can automate many sales tasks, but require extensive calibration and some risk tolerance. It’s not yet an end-to-end replacement for human reps.

2. Best fit for B2B companies with large, digitally-represented target markets. Less effective for white-glove enterprise sales or hard-to-find buyers.

3. Current sweet spot: Email campaigns. AI agents can accelerate other outbound channels too, but you may get a cease-and-desist (or two).

4. Key considerations: expect a 3-4 week ramp-up period. Managing email deliverability is crucial. Personalization may not be the holy grail you think it is.

If you have any additional insights, stories, or feedback to share, let us know. Otherwise, let’s dive in.

II. Agentic Sales AI: The Promise

The hype of Agentic AI is at epic proportions today. It’s supposed to be this super-smart barely-needs-supervision star performer. Sounds great for outbound sales, right? We’re not quite there yet. Today’s AI sales agents are more like automations on steroids.

But it could still be hugely valuable. According to HubSpot, only 33% of a sales rep's time is spent actively selling. That means 66% of the time they are searching for the perfect gif response to send on Slack.

What if you could remove the mundanity of the job and get your sales folks selling more? What about (gasp) firing some and saving runway?

Companies are certainly interested in solutions. One AI sales startup has a month-long waitlist just to get a demo. Another isn’t doing outbound sales because the founder has 10 inbound calls a day.

But will AI sales agents "make it rain"? Not so fast.

Yuriy Zaremba, CEO of AiSDR, says: "Something that I am very cautious about and I think many AI companies are doing wrong, is over-promising and over-selling. Like, oh, we can 10x your pipeline. It's just not true."

So here’s a breakdown of where AI sales can and cannot be applied today.

III. Separating the Wheat from the AI Chaff

Before you go all-in on a robot sales team, here are the AI sales limitations you need to know. From GenAI weaknesses to legal landmines, spam nightmares to shoddy personalization, I’ll unpack the current constraints and watch-outs.

End-to-end Agentic Sales AI is still a myth.

Today’s AI sales agents can send messages, book meetings, and reply to questions, but they are not at the level of Agentic AI.

They struggle with things like handling objections, adapting to new situations, or understanding context that is not publicly available.

Promises of AI identifying buying signals is great in theory but poor in practice. If the buying signal is obvious, you’re probably already too late (like when you get a shoe ad for the shoe you just bought).

Doing demos, negotiating terms, and closing deals are all still firmly in the human domain (not because AI can’t theoretically do this, but because of human expectations).

It’s best to think of AI sales agents today as highly-configurable email automations with a GenAI layer plugged on top.

GenAI still sounds like…GenAI.

CEO of Artisan AI, Jaspar Carmichael-Jack, says: "Trying to get an AI to write a good email is probably the hardest thing that we have had to do in the company....they all sound the same if you don't coerce it.”

And even with the best prompting, expect some awkward moments.

Like when your bot personalizes an email using a LinkedIn post about a death in the family: "I noticed the recent passing of your mother. May I interest you in our risk-assessment tool for anticipating life's uncertainties?" Oof (this scenario actually happened).

With AI agents, you will win some, and you will offend some.

Most AI sales agent companies have an onboarding process that customizes your emails with pain points and your solution’s benefits. But, the draft emails still require human tweaking to sound less bot-ish.

There are shifting legal boundaries.

Technically, per LinkedIn's T&Cs, you cannot use bots to send messages. I know, seems cruel - all your leads are there, waiting to be approached en masse! But it's illegal for now.

No one has been outright sued yet, but word on the street is VCs are wary of funding startups that toe this line. And LinkedIn is handing out cease-and-desists like candy. They prefer you use LinkedIn Navigator (obvi).

You can get away with LinkedIn message automation for a bit, but don't expect it to last. The good news is that AI can tee everything up for you, but you still need to (or should) click send. Scraping is also technically against their user agreement.

Issues also exist for robocalling. After "Biden" called voters and told them to stay home in the New Hampshire primaries, the FCC made robocalls a felony. Robocalls are now on par with manufacturing meth in your basement.

So where does that leave your scaled up AI sales? Primarily stuck in the doldrums of email.

But be careful to not use an AI-generated persona in cold emails - this could violate FTC rules against deceptive practices.

Being labeled a spam artist.

Email deliverability is tricky even without AI-level volume.

To make sure your emails are getting delivered, the current AI sales companies suggest you:

• Spin up lookalike domains

• Create several mailboxes per domain

• Limit sending to 30-50 emails a day per mailbox

• Take 3-4 weeks to “warm-up” your domains, where you progressively increase the number of emails sent per day.

AI Sales Agents can do this for you but there is a high risk of being spam-blocked even if you’re all warmed up.

Why? Because most email platforms were built when Russian bots were sending 80 billion spam emails a day. These platforms have invested hundreds of millions of dollars in preventing phishing, malware, scams, and fraud from reaching inboxes.

The problem is that AI-generated email campaigns can look suspiciously like those of Russian bot operations.

The secret sauce isn’t clear, but common red flags include:

• Too many Google workspaces being accessed from one location

• Appearing to impersonate a company

• Sending too many emails from one email

• Having a new domain (so ageist)

The jury is still out on whether “warming up” your domains will protect you. And even if your domains are warm, we’ve all been trained not to click on things from “similar-sounding” domains.

As Benny Rubin, CEO of Senders, politely put it a viral LinkedIn post: "Stop poking the bear…Google does not want you there. They are flagging, disconnecting, officially banning and shadow banning senders who are abusing their systems and it will only escalate."

If you're flying solo with your own automations, definitely don't tread lightly. It doesn't matter how amazing your emails are if they end up in spam (or worse, get you shadow banned.)

It may go without saying, but be careful not to piss people off with your emails. Google requires you to keep your marked-as-spam rate below 0.3% to stay in good standing.

AI-personalization can be a trap.

The conventional wisdom on email outbound is that you need to nail 4 things: right target buyer, timeliness, offer relevance, and personalization.

As your inbox will tell you, there’s an abundance of bad personalization happening. “From your past experience as CEO of BlahBlahBlah, your journey has been impressive…”.

Spoiler alert: your target knows their own bio. This type of personalization proves nothing beyond your ability to scrape their LinkedIn.

Benny Rubin, CEO of Senders, told me that less than 2% of his clients are personalizing cold emails with AI. They don’t need to. They care more about nailing offer relevance and whether the email has actually been delivered. To be clear, these aren’t luddites - they are venture-backed growth companies.

Benny’s advice: “Instead of faux-friendly personalization, a better focus would be relevance - AI selecting the features or case studies that are most applicable to the target’s business.”

Theoretically AI can do this, but it isn’t a common feature yet. My bet is it will emerge over time.

In short: AI can help with building a high-quality email machine, but don’t get too swept up in the buddy-buddy personalization trend.

An example of a digital wasteland of AI emails gone wrong:

AI sales agents aren't for everyone.

If you only have 20 potential customers in the world (and one contract would make you insanely rich), maybe don’t AI-ify your sales.

It is challenging to identify the right buyers in an enterprise that has 700k employees. AI won’t help with this. You can use platforms like Clay to mine insights, but buyer names at a massive company are not typically internet-accessible.

Some folks are using AI sales agents to "spray and pray." But you’ll likely get blocked permanently by the security gateway faster than you’ll find your buyer. As someone whose inbox at a Fortune500 has been a victim of this, I don’t recommend this strategy.

Some AI sales agent companies (like AnyBiz) use tricks like name-dropping to get to the right buyer: “so-and-so said you were the right contact.” This can be effective.

Assuming you do find the right buyer, AI cannot manage multiple stakeholders and organizational dynamics. Nor can AI build relationships, drive alignment, negotiate tricky deals, or play golf. Human reps are still best here.

And, conversely, if your target audience is very small, you'll face challenges. If caravan resellers in Tallahassee are your target audience, then there likely isn’t enough data to find leads effectively.

People still want to buy from people.

While AI seems to be nailing the AI girlfriend niche, it still leaves a lot to be desired with business relationships.

I spoke with the Head of Growth at Puzzle, a startup accounting firm that is not yet using AI for sales. While he’s excited about the prospect of AI doing most of his job, he believes two key areas will remain in the human domain:

• Even with product-led growth, many customers crave face time before forking over a credit card. Sometimes prospects will pretend to have questions just to get a call scheduled.

• Customers want their complex questions answered by a human with a degree. Rather than AI, which still counts 5 characters in the word "IRS.”

AI can start conversations, but it falls short at developing the trusted, consultative relationships needed to close deals. Especially if you’re selling, then building (also known as the fake-it-till-you-make-it sales strategy).

III. Where AI is getting a Gold Star in Sales

Jaspar Carmichael-Jack, founder of Artisan AI, is bullish on AI for sales: “The whole role of a BDR can be 95% automated with current technology.”

While he radiates founder-optimism, he’s not entirely wrong. Here is where AI excels today:

• Personalizing outreach at scale.

  • AI can craft relevant emails based on a prospect's profile, activity, or company news. This enables “compelling” messaging to more leads.

  • Machine learning models (like the one built by startup AnyBiz) use aggregated metadata on targets to enhance the chances of engagement. Things like: your topics of interest, preferred message length, and the time of day you most open marketing emails. Creepy, but can be effective.

  • AI sales platforms can also be used to iterate messaging with different audiences if you haven’t landed product-market fit yet. It’s a more direct feedback loop than ads.

• Persistence and lead nurturing.

  • Eighty percent of sales require 5+ follow-ups to close. Yet 48% of salespeople never make a follow up attempt according to InsideSales. I assume this is because we're preserving our body budget for tonight's Netflix binge.

  • AI can automate simple follow-up and even drip the AIDA sales process for you (attention, interest, desire, and action).

• Qualifying and responding to inbounds.

  • The odds of reaching a lead decreases 400x after 24 hours. Whomp whomp. AI can isolate which website visitors are your ICP and execute timely engagement.

• Automating the manual “stuff”.

  • Mainly: booking sales demos, updating that pesky CRM, researching companies and prospects.

So in summary, the reality is that AI sales agents are best used as a force multiplier for human reps today. They can offload time-consuming tasks in the early stages of sales: lead identification, scaled-up outreach, and meeting scheduling. This allows reps to focus their time on high-value activities like developing relationships, leading sales conversations, and closing deals.

Here is a handy cheat sheet in case you didn’t read anything I wrote:

So who are the leading startups who think they can solve this meaty challenge? Let’s meet the members of the AI Sales Agents Startup Club.

V. The AI Sales Agent Startup Club

I spoke with the CEOs of AiSDR and Artisan Ai, and tried out Clay.com - a fan favorite on LinkedIn - to get a pulse on what's real and what's deck fluff.

Here are their thoughts on the state of AI sales agents today, feature highlights, and my reviews.

1. AiSDR - So Humble, It Won't Even Brag

AiSDR is the highest-rated AI SDR on G2’s product review site (keyword matching for the win!). The product "works," and there are raving reviews about their customer support.

But noticeably absent from their website is the term "Agentic AI." I spoke with their CEO, Yuriy Zaremba, to understand why they aren’t trying to make VCs salivate with the latest buzzword.

As Yuriy describes it: "AISDRs are oversold a lot. And it’s very important to keep people realistic about what’s happening.” Many of Yuriy’s customers come to him complaining that they’ve tried multiple solutions already and were disappointed. So he’s focused on setting realistic expectations.

Yuriy is a YC alum with a multi-million dollar exit, after which he spent a few years doing founder-led sales at the acquirer. Hiring SDRs, firing SDRs, and generally sweating as he got the company to $100M in ARR.

He knew there could be something better than spending his days following up with leads. So he spun up AiSDR and went through YC’s gauntlet a second time in the summer of 2023. Ten months later, early adopters are seeing solid results:

• SaaS compliance platform: 70% email open rate, 9-17% reply rate, 3.7% conversions

• Investor intelligence platform: 1.8x more revenue closed

AiSDR’s Superpowers Today:

• Accesses lead databases with 700M+ contacts and built-in email verification

• Personalizes messaging based on personas, LinkedIn profiles, job posts, and company web search results

• Uses an embedded LLM step to verify brand alignment before messages are sent

• Handles incoming emails automatically in <10 minutes

• Allows you to evaluate campaign effectiveness in a dashboard

• Warms up mailboxes; automatically monitors health

• Integrates with HubSpot, Calendly, and Slack

My Review of AiSDR:

Not the flashiest option, but it’s nice that their customer support reps are integrated into your Slack. They will even manually customize your draft emails to make sure they sound extra-human.

AiSDR is best for mid-market targets. I wasn’t able to find a large-enough target audience when searching for early-stage founders. Most of their leads are from Apollo.io, but they further refine this down to only include verified email addresses.

Limitations include prioritizing leads for you, responding to inbounds, and handling objections. Their roadmap is ambitious, however. Personally, I appreciate that the CEO has a legal background and is focused on keeping it real.

[This screenshot contains dummy data]

2. Artisan AI - The YC Workaholic

Artisan AI just closed a $7.3M seed round post-YC, with luminaries like early AirBnB investor Oliver Jung. They're building AI-powered digital workers for everything - from sales to marketing to finance.

Their flagship product, Ava, is an AI sales rep with resting-robot face that will handle “95% of the sales development reps’ job.” I spoke with their founder, Jaspar Carmichael-Jack, to find out what he’s smoking.

He came up with the idea while running a marketing agency and suffering from “SaaS product fatigue” and “work fatigue.” Too many tools and too much clicking.

His experience is clearly reflected in Artisan’s ambitions: full stack and full automation.

"We want to build flagships for every single product across the sales cycle. It's a big, big piece to chew, but I think we can get it right…Right now, I feel like everyone's tech stack is a mess. It's a nightmare. We have like 50 different SaaS products.”

Ava adopters tend to see a positive response rate 0.3-3% of the time - solid potential. But there is wild fluctuation depending on your messaging maturity and the availability of your audience online.

Ava’s Superpowers Today:

• Accesses 5+ data sources (265M contacts and growing) with automated email verification

• Personalizes messages based on job title, industry, company size, and country

• Allows coaching on style for different campaigns

• Allows you to evaluate campaign effectiveness in a dashboard

• Warms up mailboxes; automatically monitors health

• Integrates with HubSpot, Salesforce, and Slack

My Review of Artisan:

Artisan has a nice onboarding flow. Ava interviews you on your value prop, buyer’s pain points, and target audience, then generates a campaign for you. Overall, the user interface is slick and incredibly easy to use and set-up.

They provide 3 neat email playbooks to pick from - Classic Sales Pitch, Uno Reverse, and Ultrahuman.

You can also give Ava notes on how to modify the drafts, though this had mixed results. I failed to get Ava to be sarcastic: “Vinay, I’ve been following your progress and it’s super sarcastic.” Not ideal. Given LLM costs, Artisan is not using the latest and greatest LLMs.

Today, limitations also exist on Ava prioritizing leads for you, handling objections independently, or responding to inbounds. They do have a very ambitious roadmap, however.

[This screenshot contains dummy data]

3. Clay - The DIY Excel Nerd

Clay.com (the data one, not the CRM) is on fire on LinkedIn thanks to a killer user-generated content growth engine. Yes, you heard me - UGC for a data enrichment company…stranger things have never happened.

While they are not strictly an AI sales agent tool, folks are using it to jerry-rig something similar.

Clay got Product Hunt reviews like: “man that's sick, meeting all my needs and expectations for such a tool if not exceeding that."

To help you find your target audience, Clay use a handy waterfall approach to cycle through databases until it hits contact gold. Then, you can enrich the list with things like: job postings, LinkedIn posts, pricing, company controversies, and much more.

The output? A simple, but thorough spreadsheet. You can also use it to create AI personalized emails directly in Clay (picking your AI model of choice).

As one G2 reviewer put it: "If Google Sheets and Zapier had a baby specialized in outbound sales."

Clay.com Superpowers Today:

• Allows you to query 50+ data providers to get the info you need (info on 31M companies and 770M people)

• Allows for personalized outreach messages using DIY In-flow AI prompting

• Enriches lead lists; Claygent (their AI tool) can even do research for you

• Integrates with many CRMs

My Review of Clay:

It’s got a bit of a learning curve and requires a lot of trial and error to set the parameters right.

A whole industry is spinning up of Clay experts who will get fancy with customizing your Clay. Take this as an indicator of how time-consuming it can be to get it right. You will also need top-quality prompting skills to generate good email copy - it's all DIY.

The spreadsheet format can be nice and familiar, but also frustrating. No one ever wants to read a whole 10K plopped into a cell.

The coolest thing about Clay is its integrations, data sources, and scraping abilities. The internet really feels like your oyster.

But without extensive setup and calibration, Clay is best used for company research or meeting prep, not as an AI sales agent. (Some Clay fans will probably want to fight me for this take).

VI. Choosing your AI Sales Sidekick

If you aren’t using AI in your sales process yet, the good news is that you’re not behind. According to Hubspot, only 37% of sales teams are using AI right now.

And Hubspot - according to their SVP of Marketing, Kieran Flanagan - has seen a conversion uplift of ~70% from using AI to thoughtfully personalize emails.

So you may be wondering: where should I start and how do I pick a provider?

As a highly-paid advisor, I came up with a sample 6-step approach (worth $300k - yours for free!):

1. Itemize the drudgery that makes your sales team cry

2. Identify the stuff that your team should be doing but isn’t (like following-up to inbounds on weekends)

3. Stack rank all the tasks from step 1 and 2 based on impact to your bottom line

4. Evaluate AI sales providers, using the following questions:

   • Does their existing feature set cover our top 3-5 stack-ranked tasks?

   • Are samples of their emails any good? (Would you respond or passively-aggressively hit “mark as spam”?)

   • How do they handle protecting my domain from being put in spam purgatory? What about legal trouble?

   • What’s on their roadmap for the next 6-12 months and are those things I want/need?

   • How do they integrate with my existing tech stack?

5. Define what success would look like (what would make you happier than status quo)

6. After implementing, continuously evaluate performance against metrics

Expect onboarding to take 1-2 hours to get into the system and at least a few weeks of tweaking, warming, and scaling to start to see results.

Many a company are hacking together their own automations, and this is an option as well. Just make sure you address deliverability issues. You'll also want to approve outputs and have visibility into results, so you can adjust accordingly.

The nice thing about the agent platforms is that they’ve rolled this all up for you.

VII. What’s on the horizon for AI Sales Agents?

As I pondered the impending assault on my inbox, I asked the CEOs what happens next. When email becomes a wasteland of perfectly-crafted messages, will it still be a useful channel?

In the short term, companies may be forced to use AI-level volume just to keep up with competitors. And cold email has shown remarkable staying power.

Benny Rubin’s thoughts: "When we started doing [cold email] in 2014, 2015 for my businesses, we always thought we had six months to go, 12 months….I think that the reason why cold email works and continues to work is it’s fundamentally a self-introduction email. There's nothing wrong with that.”

But Jaspar Carmichael-Jack warns: "Outbound is like a ticking time bomb. If anyone can do unlimited outbound at the drop of a hat, it's just not a realistic channel for anyone because they're all flooded.”

And where there are bots, there are dreams of more bots.

All the CEOs I spoke with imagine a not-so-distant future where seller bots are sending offers to buyer bots, who qualify their relevance. This could change the sales landscape entirely.

Or we may see a counterintuitive trend: a return to manual, high-touch outreach. You know, people actually talking to people.

Even that will likely look different than today. Yuriy Zaremba says real-time AI-voice modulation for calls will emerge, as long as robocalls are still banned. Yay. Can’t wait to hear a less-sarcastic version of me.

Whether we're heading for a bot-powered buying fest or a return to old-school networking, one thing holds true: better be ahead of the game than behind.

VIII. Key Takeaways

In summary:

• AI sales agents aren’t going to get you to your founder dreams of FIRE. Great at grunt work, not so hot at the nuanced stuff and closing deals.

• Email is still the best playground for AI outreach. LinkedIn and cold calls are dicey territory - unless you want to use cease-and-desist letters as office art.

• The AI sales startup scene is hot. But remember that many of these companies are younger than your COVID sourdough starter. Approach with a healthy dose of skepticism and patience.

There you have it, folks! If you made it this far, you deserve a gold star.

Drop a comment below with your AI sales agent war stories, insights, or burning questions. And let us know if you liked this new style.