AI Agent Experiments, Lisbon Recap, Algarve, Agency Savings, Everything Breaks, Future of Outbound

Hey!

Why you’re getting this: I'm Jakob Greenfeld. I run Sales.co and this is my personal newsletter. I send this every month or so to interesting people I’ve connected with and friends I want to keep in touch with. You can unsubscribe at any time, and I won't be mad. I won't even be notified!

• Lisbon. As mentioned last time, Ryan and I met in Lisbon for our third belly-to-belly week.
  
  Met great people like Charlotte Grysolle, Luis Nicholls, played ping pong, and spent many hours at Honest Greens working

  

• Algarve. After Ryan left I took the train from Lisbon to Faro where I met with my girlfriend who flew there from Aarhus. We had booked an Airbnb in Vilamoura, which is a small town a 20-minute drive away from Faro and had an awesome time exploring the surrounding beaches.
  
  The Algarve is now one of my favorite regions on earth. Incredibly friendly people, perfect weather, beautiful nature, and great food. Might be different during the high season but at least during the off season it was perfect. We’re already looking at flights, planning our next escape from the Danish winter.

  

• Saving money. Since we’ve started to take out bookkeeping a bit more serious, a main focus has been ruthlessly cutting costs at our agency. We canceled unused subscriptions, downgraded plans (e.g. Slack), and negotiated discounts wherever we could. Many small charges quickly add up to pretty substantial sums that can be better put to use elsewhere.
  
  Black Friday was a great opportunity. I was able to get us some nice discounts for tools we’re already using simply by asking: “Hey, I saw you have this Black Friday offer. Any chance we can get that too as loyal customers?”.

• Everything breaks. Last week our main lead data provider made changes which meant we can no longer use them, one of our main email account providers changed settings which hurts deliverability big time and means we have to switch, and last but not least the main sending tool we’re using f*cked up big time and introduced a bug that caused dozens of our accounts to get (rightfully) banned and our whole account was suspended by another email account provider.
  
  So I basically had to rebuilt our complete backend in the span of 48 hours.
  
  Stressful but ultimately turned into something positive.
  
  The problems pushed me over the edge and we finally switched to a better lead data provider, better email account provider, and made the first steps of becoming less dependent on the sending tool. While things looked pretty bad for a few days, we now actually came out ahead.

• AI agents. While agency life keeps me busy I’m still trying to play with new technologies whenever I can. We test a lot of AI tools at our agency but most of the time things are not quite ready for production.
  
  For example, yes, you can have AI manage inboxes and handle replies. But a human will currently still do significantly better for anything beyond the simplest tasks.
  
  But I keep testing as the technology improves fast.
  
  One fun recent experiment is an AI agent I coded that scours the web for pain points, then rates them for their monetization potential and difficulty to solve.
  
  Not perfect but already surprisingly useful in surfacing interesting problems.
  
  The project got a shoutout from Anthony Pompliano on the My First Million podcast.

  

• Future of Outbound. One thing I keep thinking about is banner blindness. There’s no denying it applies to outbound efforts. As more cold emails get sent, the number of replies per email sent goes done.
  
  But what to do about it?
  
  An obvious answer is “send more emails”.
  
  The problem with this approach is that the number of relevant decisions makers for B2B companies is not that big. For most of our clients, we’re already making sure that their offer is top of mind for every single relevant decision maker in their target market. Increasing the sending volume literally makes no sense.
  
  Instead, the correct approach is to focus on quality.
  
  While people’s inboxes are full with cold emails, most of them are lazy and generic. So it’s actually not that hard to stand out.
  
  On the service delivery side my focus is now on 10x’ing the quality of every single campaign we launch for clients.
  
  This starts with the right targeting. There are lot of data providers but most fail at the most basic filtering tasks like “give me only b2b companies” or “only SaaS companies”. I’ve tested every single tool out there and the only solution seems to build something in-house.
  
  Similarly, for getting contact infos of relevant decision makers, most platforms use terribly outdated or wrong data. I’ve tested all providers and the difference between them is up to a factor of 4. I benchmarked providers using data from positive replies we recently generated for clients. For each of these emails I know what the correct contact infos are and then tested which provider was most successful in finding it. Might do a full writeup but by far the worst data quality in my tests has Snov.io. I’ve also explored building an in-house solution but while it performed better than most, it does not really seem worth it at this point.
  
  Sending tools are another big factor. Again, I spent a ton of time testing every single tool and really almost all of them fail at the most basic things. It really seems like most people building these tools do not really send many cold emails themselves. For example, hardly any tool makes it easy to preview the emails that go out or allow you to see which variant produces positive replies. Again, might do a writeup but for now, just let me warn you to not use Woodpecker. For example, on their pricing page it says that a certain number of “Contacted prospects” is included and only in the fine-print they mention that they count every single prospect you add to a campaign regardless if they have been contacted. Very shady.
  
  The biggest problem seems to be that every tool seems to try to do too much at once instead of just nailing the fundamentals.
  
  I’m now strongly leaning towards building a solution in-house. Sending emails using SMTP can be done with a few lines of Python code. Fetching emails using IMAP is just as easy. I’ve already started initial tests and so far things do look promising.
  
  Last but not least, we have personalizations and copywriting. We’re using in-house solutions for many months now and I still haven’t found anything better. But if you know any tool I should try, please send them my way!

  

Talk soon,

Jakob