What’s Cookin’ in Hook-land? (Vol.8)

Hey There! Below is our/my weekly newsletter on anything integration and APIs. Fun tip- any reply to this email goes straight to me, and I'll respond! On a personal note... I'm a day late with this one, but I swear I have good reasons. This past weekend, we did something that we hope becomes an annual tradition - a long weekend/vacation with three other families. Eight kids, all 4 and under, in relatively tight quarters. Let me just say... if you're going to attempt this kind of gathering, make sure you have a yard and some structured activities planned. Otherwise, you're basically running a very chaotic, very loud daycare. But here's what made it special: having close friends who are in the exact same life moment as us. There's something refreshing about conversations that aren't led by professional expertise or networking. Instead, it's "hey, can you grab that kid before they fall off the couch?" and "whose turn is it to make lunch for the small humans?" It reminded me of my parents' friend group from college - they've been doing annual gatherings for 40+ years now. The closeness they've maintained has led to healthier families, healthier relationships, and honestly just more fun as they've gotten older. I'm hoping we can replicate that for our kids. The other interesting thing about the weekend was trying to translate what I do day-to-day to people in completely different industries and roles. It's a good exercise in making integration work accessible beyond the tech bubble. Turns out explaining APIs to a teacher and a marketing manager forces you to get really clear about why any of this matters. Speaking of which... Best, ~Sean What we're seeing All products are integrations (and Freckle proves the point) I saw a LinkedIn post today from a startup called Freckle that got me thinking (once again) about something I've been mulling over lately: all products are essentially collections of integrations, just packaged differently. If you haven't heard of Freckle, take a look - it's essentially a simpler version of Clay with a more specific workflow focus. That's a side note on what really caught my eye about the post, however. The post was helpful in sharing all the tools they use in their stack. What's fascinating is that as I read through their list, several tools they mentioned are purely ecosystem-driven value propositions. The line that really stood out: they were using Attio (an AI-native CRM) but switched to HubSpot because "HubSpot's ecosystem of pre-built integrations made us feel comfortable scaling, even though we admit Attio's UX is better." Think about that. Superior UX wasn't enough. They acknowledged the competitor had better user experience but still switched because of ecosystem breadth. There are two or three tools in their stack where the decision came down to integrations that work well and can grow with the customer as they scale. Then there was the Calendly mention: "integrates with everything" as a core value prop. Not "we have the best calendar interface" or "we have the most features." Just: "we play nice with everything else you're already using." The lesson? You can build a superior product, but if you're not thinking about your ecosystem and making sure those integrations work well, you'll lose users even as they grow. You'd think growing customers would grow with you, but they actually grow out of you because of integration limitations. MCP namespace collisions are becoming a real problem Ayan Barua, CEO of Ampersand, posted about how both enterprises and individual users experimenting with MCP servers are starting to run into name conflicts. What that means: imagine installing two integrations that both try to use similar tool names or descriptions. Your AI chat gets confused about which tool to use, and performance starts to degrade because the LLM can't distinguish between them. This isn't integrations "yelling at each other" - it's the AI getting confused and performance degrading as a result. This is a problem that hasn't been solved yet because MCP is a protocol, not a normalized, centralized marketplace. Anyone can build and plug in their own MCP server (I'm currently building my own using AI tools), but there's no traffic control. For non-technical folks, here's what this means: as you add more integrations to your AI chats, clients, or agents, you're going to start running into tool overload problems. We'll start seeing people who build better integrations and better management tools for your AI to use. The growing pains are real, but they're also signs of a maturing ecosystem. The AI backlash is real, but so is the daily utility There's been significant backlash about AI lately - complaints about failed enterprise AI experiments, and the general sense that nobody's solved the "grounding in truth" problem with AI systems. Some of this backlash is valid - there are real security concerns and overinflated valuations. Some critics are treating it like VR or 3D printing - a fad that will fade. The criticism isn't about GPT-5 delays, but rather that GPT-5 performance and output is fundamentally just another iteration on an LLM. There's no pivotal AGI moment or transformative leap - it's incremental progress. But here's my take: I'm reading the macro analysis and anti-hype stuff, and then I'm experiencing daily utility that's genuinely game-changing. The code I get when I prompt an LLM and review it is basically what I was looking for. I can ask for something in a programming language I've never used before and get working results. I can throw together an HTML page for a quick prototype. I can take a proposal template, share my meeting transcript, and say "pull out what I'm supposed to deliver here and update this work order" - and it works well. I think I agree with analysts who say it'll take 10-20 years to fully unpack what's been unlocked with LLM technology. We're not near AGI, but we're also not in a bubble that's going to pop and disappear. My approach: keep leveraging the things I find valuable, keep evaluating them for utility, but don't bank on them behaving the same way day after day or year after year. What's your take? I'd love to know where you land on the spectrum between "it's all hype" and "it's transformative." What we're hearing When embeddable iPaaS doesn't deliver, the services vs product question gets real Had an interesting conversation with a client this week who had a frustrating experience with an embeddable iPaaS tool that didn't meet their expectations. As we talked through their options going forward, it highlighted a key distinction that I think is worth calling out. When you buy an integration platform/embeddable iPaaS, you're buying a product. When that product doesn't solve your specific needs, they generally can't flex outside their platform to help you. They don't have a model for custom solutions, and unless they're a quasi-services firm, they can't promise services to resolve edge cases. When you work with a services company like us, you're buying people to achieve an outcome. People who can flex to different tools and find whatever will actually work. This is why I'm feeling pretty bullish about Left Hook's positioning. We're not trying to give you a product to DIY with. We're the plumber, not the plumbing. The person selling you a sink isn't the right person to be frustrated with when it turns out you needed a custom-fit bathtub. But they're also not well-equipped to help you with the "custom fit" part. If they sold you a sink promising it could be made into a custom-fit bathtub, that's on them for not understanding your needs. Which brings me back to the value of expertise. Feeling pretty good about that positioning. Integration fatigue is real (and understandable) Another theme that came up in multiple conversations this week: integration fatigue. People get excited about the potential value of an integration. When you see it working live, it's genuinely exciting! But for product teams, value is only achieved when users are actively using it and it's clearly providing outcomes. For end users doing integrations DIY (especially those experimenting with AI tooling), there are definitely "AHA!" moments that feel transformative. But the real value is in using those integrations and automations consistently, then tweaking them over time. Sometimes they fade into the background and become a non-problematic part of your workflow. But more often, they break, get brittle, or the initial promise was exciting but only delivered value a few times. Even on the home front - I love that Alexa can talk to my TV and lights, but it's incredibly frustrating when it fails to understand or connect for some random reason. In all cases... enter the fatigue. It makes you want to give up on the automation effort altogether and move back to things that "just work." But here's the thing: to get to that "just works" point, someone has to do the unglamorous work. Diligent updates. Getting feedback. Reviewing error logs to determine what's a real error versus noise that should be handled gracefully. Finding and fixing pain points. As someone said to me recently: "It's boring and painful work, but it's critical and valuable." That came up in a few conversations this week, and I want you to know - I feel it too. And we see you working through it. You got this. What we're doing HubSpot UI extensions: fascinating architecture, frustrating limitations We've been deep in the weeds working with HubSpot's UI extensions, and for anyone technical, it's really fascinating how they've architected it. On the surface, it looks like you can just build React components and upload them. But when you look under the hood, they're not true React components. They're essentially representations of React components. You build it as though you're building React, but when you upload it, HubSpot decomposes and recomposes it so they can verify it only uses code they condone. This means they restrict all the UI components that are available. You can't just add another component to the mix or combine a few creatively. It's a little restrictive, and you can't get as creative as you might want, but at least it's familiar enough - you're actually building in React, just with a different build and compile process. The feedback: there's a limited set of things you can do in the UI. Translating the desired outcome and UI to what's actually possible is kind of an art at this point, with some knowledge and skill involved. But it's cool that they've at least expanded the abilities for integrations beyond what used to exist. Now if only we could get them to let us do more sophisticated UI components and transitions, we'd have a really fun thing to play around with. Revamping our offerings (and remembering automation is still very much needed) On the organizational side, we're looking through and revamping our offerings. That includes the unlimited integrations offering we've mentioned - once we get past this initial cohort, we'll probably set the floor price at $10k/month. When you break that down, it's actually super high value for less than the cost of an engineer, given how we can reproduce our expertise across many different clients with the right monitoring and maintenance tools. We're also opening back up our small business automation approach, which has led to some fascinating questions about whether businesses still need automation in the age of AI. Here's the thing: I'm working with new AI and LLM technology, but I'm still talking to people who use Excel every day, jump on phone calls for everything, or maybe tried ChatGPT once but haven't figured out how to incorporate it into their daily life. Construction workers who aren't about to open dual monitors and figure out 3D CAD design for their next project. There's still a massive need for automation and integration in the world, and that need is relative to the outcomes people get. The fundamental jobs haven't really changed much beyond some roles inside the technology industry specifically. We just have a few extra tools at our disposal. Two tracks: managed integration services + business process automation What we've settled on are two main offerings: Managed Integration Services - We help with integration platform connectors and direct native integrations leveraging any technology (Frigg, embeddable iPaaS, or custom code), then support and iterate over time. This aligns to your business outcomes and KPIs. Business Process Automation (for SMBs and end users) - Similar to what we've done before, but now including AI in those workflows. Our entry point is when people hit the limitations of self-serve integrations. Our approach: shadow and identify low-hanging fruit pain points, then help transform businesses with available tooling. Most likely, they get set up on platforms or direct integrations within a given tool, with some guidance so they have a process and approach rather than a quick fix. The goal is to take away some of that integration fatigue and help people grow their business or reduce strain so they can focus on what they actually want to do. Contemplating AI coaching and implementation assistance We're also thinking about offering AI coaching - helping with assistant workflows, MCP servers, figuring out the right stack, and general consulting rather than preset packages. The concept: here are some tools and workflows we could augment, and we could run coaching for people in groups within their companies. Think mini-training followed by "let's get you onboarded into using AI effectively" so you avoid the "we tried it, it was neat, but didn't really work" burnout. How can you be effective with AI inside whatever you're doing? That's the question we want to help answer. I'd love to shape this offering with anyone who has questions or can imagine what might work for them. We're still looking for companies to take us up on our unlimited integrations offer. A few spots left if you're interested in moving fast on integration strategy. Reply to let me know your thoughts. What do you want to know more of? Smash that unsubscribe button below if you don't like it. Forward to a friend if you do! Until next time! Sean Matthews Co-Founder Left Hook