2nd Week May 2026 - AI SaaS Market, Open Source AI Agents

Speaker A: Hi and welcome to the Cognicore Consulting podcast with me your host, Dr. Samuel Zerubba Smith, where we provide you all the news on things impacting tech and consulting. And welcome to the Cognicorp podcast for the second week of May 2026. Our news story today covers a variety of research that comes out regarding small and medium sized businesses and enterprise choices regarding AI agents. We'll be looking at a number of reports coming out on the top AI agent software packages available to both small and enterprise customers. Our Benchmark testing will look at reports around benchmark testing for a number of different providers. We'll look at particularly first is the horizontal builder category. This is the build your own AI agent category. News reports have come out recently on the top AI agent builder providers including Lindi, Zapier Agents, Make.com, relevance, AI N8N and Gumloop. We'll look at reports across these various providers within the AI agent software, uh industry and other competitors as well. The first report comes out regarding the builder category. These are the horizontal builders that uh, are sort of that classic style of AI agent solution where you don't write code as the user but instead you talk to an AI or you use some kind of drag and drop style workflow to build your AI agents and build your automation. The first company we'll be looking at today is a company that's had a lot of recent uh, media, press and news stories and articles coming out about them, which is Lindy. Lindy is the no code conversational AA agent specialist. Lindy excels at email triage, CRM updates or customer relationship management, updating, scheduling, uh, both for you and your employees and internal support tasks. Multiple independent reviews in the studies we looked at for the beginning of the year of 2026 rank it arguably the best overall SMB provider for conversational agents. So if you are a small or medium sized business or you're an enterprise looking to adopt uh, early uh, uh, leaders in the AI agent space. Lindy is one of those major leaders. Lindy provides out of Box SoC2 and HIPAA compliance which can matter a lot what we see for folks in healthcare, legal or financial services. And the freely provided template library for LINDE users is one of the strongest template libraries we see in the space today. Which means most business use cases have a good starting point uh for working from day one which decreases time to get it up and running where Lindy often falls short or which much of the research indicates of the downfalls of Linde currently in this kind of spring, summer time period of 2026. Currently the integration footprint for Lindy is much smaller by a wide margin compared to competitors. Meaning that uh, there are less ability for Lindy to work with other say major cloud providers. Um you know you have AWS or Azure or maybe you're already looking at other you know companies uh as well like Zapier or things like that. Um so the issue of uh having deep automation uh complexity with other major providers again your azures, your AWS's, your GCPS and Google et cetera, Salesforce et cetera. Um Lindy does fall short in this category in most of the research reports coming out about it. Um the other one that is mentioned commonly in the same reports is that the price scales steeply once you cross from pro to team at ah the kind of the company level side. So if you go up in terms of pricing so Pro starts at $20 a month for Linde teams around $70 a month and company $103 a month. Um, the scaling you're getting in terms of the pricing is said to be uh, not great. There's a number of reports out there about how it's not quite the most efficient use of funding uh once you can hit those higher tier rates and things like that. Particularly if you look at for teams running AI agents in parallel. Lindy uh is quite expensive for running multiple AI agents. For example at seen um Lindy is also cloud only which means if your data has to stay on your own infrastructure or you're in a regulated environment, um, that can be a deal breaker for many folks as well, many companies as well. So when would you choose Lindy? Well if you have a small technical team or you have a small business team um so it's great for both technical and non technical teams in that sense. But um, you have to have a small team. You want a standard kind of software as a service style stack. Um, you may want conversational AI that can handle email, handle your calendar, handle some CRM tasks without writing any kind of code. And again if you do care about SOC 2 or HIPAA compliance, um, this company does work very hard to provide services that meet that kind of compliance um when you should not choose Lindy. So if you want full ownership of your data of course it is cloud only Lindy so it does not have any on prem ah abilities. Um additionally if you're doing looking at multi agent collaboration or many multi agent groups working together, many different AI agents working together, uh it does not excel at multi agent work. It's more Designed for having individual agents do a task one by one, but not having two AIs talk to each other for example or you know, definitely not having more than two or three agents talking to each other. So if you're looking for one AI agent that can automate all of email, or one AI agent that can automate all of uh, a customer outreach or things like this, or CRM, um, or your calendar, you could have a single agent developed for that use case very, very well. That's kind of what this main bread and butter of this provider is. But for multi agent collaboratory uses, say like research or any kind of um, you know, where you have multiple agents as any governing and talking to each other. It does have some negative reviews and negative benchmarks uh, at that particular level within the research. The second uh, um, error we saw a lot of research on and a lot of uh, news articles coming about uh, for this week is on the company Zapier and their AI, uh solutions as well. Um, Zapier, the pricing on it is roughly you get 100 free tasks you can use per month at the free level. Um, there's a professional level at $20 a month and a teams AH level at $70 a month with the company level coming in at $105 a month. What is Zapier's main advantage? Well Zapier's main advantage is how many integrations it has. It has over 8,000 different software integrations, meaning that just about any other uh, product service you're working on at your firm, there is a Zapier integration for it. So if you're on Google Cloud or Google Workspaces, you're on aws, you're on Azure, you're on Slack, you're on Salesforce, you're a Notion or HubSpot or any of the other major providers out there. Um, Zapier, you know, definitely wants, definitely uh, has the ability to, to utilize resources from all those providers and then use Zapier AI to talk to them or to create automation with them and things like that. So it's primarily structural advantage, it's primary advantage towards competitors. Um, you know it scores very highly in the research regarding how executives uh, view it in terms of use cases for its uh, uh, integrations and it's the, the wide variety of integrations that it offers. Um, so essentially what the Zapier AI Agents product is is it's a conversational AI agent layer or a generative AI agent layer over an existing ZAP software engine which then has a variety of other kind of task libraries of templates and things like this to use on, even on top of that. So it's mature in terms of its ability to integrate with other major providers in the software space. Um, the negative reports we see coming out around Zapier or the ah, reviews and the research in terms of polling from um, managers and executives who provide feedback uh, in the research is that um, complex agent reasoning is somewhat weaker than they expected compared to a competitor's. Um and the Zapier platform is less AI native than some of the other providers. Um the original Zap engine was a more traditional software engine uh AI Some, some executives argue it feels more bolted on like they just essentially took other AI models and place it on top of their already existing software automation stack that they were already selling prior to using Zapiera agents. Um, so there is a issue of not maturity in terms of integration of course because there's many different softwares that integrate well with it plots integrate with it but in terms of its maturity regarding AI native or using complex AI models kind of cutting edge AI models that the AI models don't feel as cutting edge. Um with Zapier they feel a little more like they've used some open source models maybe bolted on top of um, a more traditional uh, software automation workflow or that the models themselves don't feel as custom or as uh, a cutting edge as some of the other providers do and things like that. Um, additionally uh, if you already are paying for a Microsoft subscription um, or some of the other major uh, um, you know providers such as Salesforce with their Agent Force or Microsoft has Copilot Studio for example or AWS and Google also have similar solutions that are coming out currently. Um, there's little incremental advantage because your current license probably already covers most of the things that Zapier AI agents do using those big platforms. So yes it integrates very well with AWS and Azure and GCP and Salesforce et cetera. But if you are heavily dependent on those providers already um, those providers may already be providing you something that's similar to what Zapier does in terms of having an AI agent layer of software essentially built on top of uh, maybe more standard layer of software automation uh, stack. So uh, you know, in terms of the actual benefit to you as a customer, do double check to make sure that your current provider you're not already using like one of the big cloud providers like a Microsoft or an AWS or a Salesforce, um, doesn't already provide something very similar at a lower Cost that you're already essentially paying for as part of your subscription to those other services. Um, when you should be choosing Zapier would be things like you want a more standard SaaS style, ah, subscription, um, that's not through a big provider like a Microsoft or a Salesforce or an aws and you want the broadest kind of integration level with third party libraries or third party apps. So if you're using a lot of different third party apps or using your Slacks or your Notions or Hubspots or you've got a lot of other custom, maybe legacy type systems or things like that, um, that are tied to a variety of different subscribers, um, and you want to build like a custom link between all of them or custom automation between all of them. That's where Zapier can really be, you know, really stand out. Um, and so you know there's already a lot of Zapier customers who are using for you know, just traditional software automation, non AI related automation. And that's what you know they were doing already for a very long time. So now they're essentially added the AI agent component to it or on top of it which again it gives it maybe that bolt on feel or that, that less AI native, that less cutting edge feel. Um, but it has so many integrations that can be highly useful if you're already doing a lot of other work with other, other parts of the ecosystem. Um, and again you're not just a purely Microsoft or you know, Salesforce shop or things like that which already have a lot of these things built in to compete with Zapier directly. So if you are Microsoft or Salesforce, do look at um, what's embedded already in your current ecosystems or if you you know, say you didn't need um, you know, a variety of things to be connected to, this might feel a bit overkill because it does have the ability to connect with many different solution providers out there, a lot of custom built plugins and things like that. It has you know, for, for more traditional software automation. And if you really need cutting edge artificial intelligence, you know, uh, uh, models to be used by your agents, you're doing some very complex work. Or if you need multi agent reasoning, you have agents who need to talk to each other that are in a very autonomous way to make decisions on their own. Um, you know, that kind of cutting edge AI agent idea, um, not just trigger various different workflows with you know, AI bot brought in here, AI bot brought in there. But you know, if you need a truly autonomous uh, reasoning, uh Setup with M many agents doing complex things Zapier uh may not be for you because again it uses some of the less cutting edge aspects in terms of the models themselves and is not truly a multi agent framework. It's essentially software automation with agents bolted on on things like that. Um, the third um company that was listed in a number of the research um, that came out in the last few months um, that had a large amount of reviews and executives discussing it as a solution um, was make.com um make.com you know there's, there's a free tier uh and there's the core plan starting at only $10 a month. Um and there's of course higher level plans with Pro and things like that as well that scale and add additional AI features. Uh make.com is arguably one of the cheaper entry points um for the AI agent automation SaaS style platform category M but it's very linked to the visual workflow builder style of automation. So um, the idea here is that uh make.com provides a great interface for doing workflows automation. So you imagine connecting two websites to together two social media accounts to together you know many, many times, right? You have 20 different accounts all connected together or 20 different databases linking together with an automation in between them and things like this. Um, so that's what really make.com excels at. Um it's got this very block uh based you move blocks around it and have, have arrows and have uh, you know curved lines connecting them together which you know shows you the visualization of that automation. Um and AI is layered into these existing workflows but it's not a primary interface of the workflows. So if your team already thinks a lot in terms of how do we automate a workflow from say HR to finance or from multiple departments, you know department one, department two to department three, uh, you know with some approvals in between that or something else like that. Having that kind of workflow approach is exactly what make.com uh excels at. Um where make.com falls short is that there is little to the true uh agent reasoning. Um it's more of like a workflow automation tool to help you design how to automate various softwares talking to each other with chatbots as a type of software kind of built in there. It is not again that complex agent reasoning you may want if you have multi agent problems that need a lot of uh, say research done or things like that and having the agents talk to each other quite a lot lot. Um, you know it may essentially Frustrate you if you have one complete autonomous agentic decision making or you know completely you know hands off style AI agents that you can trust to do complex tasks or have multi step conversations with customers or clients and things like that. Um, the visual kind of canvas of moving the blocks around, moving the workflow from step one to step N, as you build that out for how complex your workflow is, that can be very powerful for uh, visualizing it, you know from, from a, from a planning standpoint. But it does have a real learning curve and it is recommended that you use training for your employees who work in that space. Um, technical automation teams that already think in terms of workflows or step by step processes. Um, make should be very good for you, right? That's like it's a good quality tool for that traditional software workflow automation. With generative AI as a tool built in. You can add AI agents, you can add an AI chatbot, but it is not AI native in that sense and things like um, it is more budget conscious again has that lower starting price point. Um and if you want essentially AI or generative AI as like one node or one, one piece of just a larger software automation it's great. But if you want complex AI agents talking to each other and doing autonomous work completely on their own or working together, it is probably not the correct tool for you. Um, at least that's according to the research. Right. Um, so if you need conversational agents to handle full support or scheduling end to end or you need some really true high level cutting edge reasoning from your gen AI rather than just kind of triggering various actions among a chain of linked events. Um, you know this may not be uh, the one that meets all of your qualifications according to the research. Uh next we have Relevance AI is another company uh, that has came out that has a large amount of um, research being being associated with in terms of gathering opinions and feedback from executives, managers and engineers. Um Relevance AI has a free tier as well. It uh, has some starting plans starting from $30 all the way up to $350 per month which is like a credit based model similar to some of the other big AI providers do for the LLMs and things like that. Um Relevance is a true specialist in the multi agent category. So unlike some of the previous M providers uh we just discussed um, Relevance AI focuses just on the multi agent specialist part of it. So agents talking together, cutting edge agents working together, scenarios where multiple bots, they collaborate on a certain workflow with say large amounts of memory needed. So large Context windows, things like this. Um, and we want large internal data processing, scoring logic, um, machine learning, custom machine learning built in real time and things like this. Um, you know that's, that's where this relevance AI really shines. It is SOC 2 compliant out of the box. So that's very good if your goal is to build, you know, what, what venture companies are now often considering like a new digital workforce where you have you know, 10 different agents all with different names. Each of them does a different thing. Um, and you want to you know, hand off one task to one agent but have the rest of the agents all talk to each other to solve that eventual problem. Um, like handing off meetings to a booking agent, which then gets sent to a scheduling agent, which then gets sent to a sales agent, which then gets sent to an engineering agent, E.T. this is where um, relevance, uh, AI is really engineered um, for where it's really excelling at this kind of multi agent pattern process. Um, where there's been some negative feedback on it in terms of the research, is that the learning curve has been considered steeper than most of its competitors. Um, so it has a time, a high technical ceiling on it, uh, and the floor of course, high technical floor as well. But it can do a lot because it's got some very complex multi agent patterns, very complex cutting edge agents. But it does take a lot of good technical knowledge to um, to get that kind of up and running. Um, the credit bait pricing model is known to create some kind of budget unpredictability. You uh, have to kind of manage your tokens, manage your outputs there just like you would a lot of the big other AI providers. So there's definitely a kind of a budget uncertainty there which can be negative to certain companies, um, who like that kind of more, more standard cadence of 30 bucks a month, 100 bucks a month, whatever it is, every month, no matter what. Um, the integration is also narrower than with some of the other competitors, namely like Zapier and things like that. Um, so you know, for, if you're looking to really link it to a lot of variety of other types of uh, uh, paid software solutions, your salesforces, your Hubspots, your slacks, etc. Um, it can do some of it, but it's not nearly as um has as many integrations as many of its competitors do. So low on the integration numbers and things like that with third party providers. Um, you know, if you're, if you're looking for only a single agent or you want to turn a database into a chatbot or you want to have some kind of more software automation tool where a number of pieces of software are automated and it ends with a chatbot. Or maybe it's only one step that involves an AI. For example, you probably don't need the complex multi agent stepping of relevance, uh, AI that's really only designed for, not only, but it's primarily designed for um, many agents working on complex tasks. So if you don't need many agents, you Maybe only need 1, 2 or you have individual, you know, bots that tackle one thing but they don't talk to other bots. This is probably an overkill in terms of this tool for you and things like that. So if you do need multiple agents specializing on a particular workflow, complex sales, operations, research, complex software development, this is where relevance AI can be very useful and things like that. But you do need that kind of technical team to manage the complexity. Um, you know, you know, in terms of, you know, having things you, you have to have, then it's that complex team you need. Maybe if you need a predictable monthly cost, this could be not necessarily a great tool for you. Or if you're not using multi agent workflows and only need a certain, you know, one agent at a time or agents don't speak to each other. This may not be the relevance AI may not be the tool for you. Okay, so uh, N8N is another um, provider within this space. There's a lot of great uh, reviews of N8N. Um, it is kind of the classic open source developer choice, um, solution. Uh, it is free if you self hosted right. So it's open source in that sense. Um, cloud starter starts around 20 bucks a month. There are higher tiers available with more executions if you want more power and things like that. But it is the kind of classic style open source solution within this space. Full data sovereignty you are, if you self host it, you own all your own sensitive data. It never leaves your infrastructure totally. You don't have to use the cloud at all. If you don't want to use the cloud, you can be entirely self hosted. It has the deepest AI capabilities. Um, so anything you can think you can want to do, you can essentially do with it. Um, it's, if you have a very technical team you can wire up extremely complex custom logic, extremely complex multi agent systems doing really interesting work with it. Um, so if you want to manage your own infrastructure, infrastructure and you want to have the most kind of cutting edge availability to you, NAN can be a great solution, um, to be looking at. There's a lot of uh, developers and managers kind of reviewing this product right now and providing feedback to researchers on it. Um, you know it does require strong technical knowledge base for your staff or your developers. So if you're looking for that kind of polished off the shelf cloud native solution, this is not it. This is not it for you. This is a very, a technical uh tool for technical users. Again it's open source. It's kind of got that classic open source dichotomy there. So if you're leading operations at a say 100 person company and you don't have a strong in house engineering team or you're a big you know enterprise but you don't want to run things on prem say for legal reasons or things like that, this would not be the tool for you. Right? This is that kind of classic. You have at least one team or one person to manage the product from uh, a complete end standpoint. Maybe if you operate in healthcare, legal or financial services or government services in some way you have to keep the data on your own infrastructure. That's where it's a very strong choice. Right? Like in that classic cloud and buy versus you know build it yourself open source style uh options. Again this is the open source source option so you want to avoid vendor lock in or any kind of long term commitments to any of these AI agent vendors I mentioned or you want to you know even avoid you know long term commitment to an open AI or anthropic or um, you know minstrel or one of these other providers out there. So you know strong, strong open source ecosystem built here that you can avoid that long term commitments, long term vendor lock in things like that. If you want overall lower lifetime costs you know in regarding to using the cloud. Of course being on prem having your own infrastructure can keep your costs lower. Um so a lot of the classic open source advantages of less commitment, you know ah, lower ah cost in terms of if you use a long term, long term horizon window. But of course it lacks that convenience you get from a provider who's, who's doing a lot of that work for you and things like it it so you should not choose this tool. Again if it's, you shouldn't probably use them in most open source tools but in particular a tool like this if your developer or staff cannot own it or you don't have the ability to pay someone to manage it yourself or you can't manage it yourself due to technical lack of abilities, um, you just need to polish templates or you need quick setups you know this is not it, right? You have to do your own templates with this, you have to do your own setup with this. You have to pay a vendor to handle your infrastructure or company. You just want to pay to manage your agents. This is not it for you. This is the more classic open source style approach of you're the developer, right? You're doing on your own infrastructure or your own cloud use cases on AWS or Azure, GCP or whatever. You manage it end to end. You can do anything with it. It can be extremely complex. It can do multi agent logic, it can do multi agent agents talking to each other. Deep, um, deep AI capabilities with cutting edge models. But because of that you do have to have that strong engineering and technical team to run it for you and things like that. And then finally, um, the last reviewing category, uh, for these AI agent research, uh, this week is Gumloop. Um, this is a new uh, uh, company. It's got free tier plans available and paid plans starting at roughly $40 a month. Um, it is a platform for natural language agent building so you can talk to it in plain English or other languages and it can create that agent for you. Um, it is, has a built in LLM access so you don't need to bring in your own API keys from OpenAI, ChatGPT, Anthrop Topic, that sort of thing, et cetera. It's got that own, it's got its own built in LLM. So that's very nice. It is open source, it works with some open source areas like model context, protocol and things like that. Um, and it has a number of integrations like Slack but not full integration. It's got some integrations right? Not, not a ton. Um, it does fall short in terms if you want many integrations like you should look at being a links app, your make.com and things like that. It has some integrations against the big providers, your anthropics, your, your slacks, maybe you know your, your azures, your GCPs, your AWS. But it doesn a lot of integration. So it's smaller in terms of the number of integrations it has. Um, a number of the reviews from the executives who are part of the research group um, said that uh, it's got some good reviews but there are only a few reviews. So it's not like you have a huge amount of data uh, to look at for reviewing this particular type of product. It's a new entrant in the market and therefore your operational risk is higher for it. Um, it's potentially a Good choice. If you have a small team or you're starting out yourself, you're a startup company or you're just looking at this from a prototyping standpoint at a big enterprise. So if you're a manager at a big enterprise or a medium sized firm, you're like, well, I want to try what's the new cutting edge out there but I don't want to commit to anything yet. That might be a good place to start with Gumloop and things like it. Um, you have to be comfortable being on a new platform that's relatively untested and you don't need a ton of third party connectors to legacy systems. You have some, right? You have your slacks, you have your aws's maybe and things like that, but you don't need, you know, more, more exotic third party connectors to integration systems to integrate with legacy systems. Right? You're okay being on the kind of more modern, cutting edge aspects of, of this, this product which is again new to the market and things like it. Um, if you need vendor stability or long term lock in or long term track records, again this is not the product for you. It's a relatively new product and so if you need that deep integration stuff, it's not going to be there. You know, it's if you're a, let's see here. If you want total ownership of your data or you know, lower overall lifetime costs. We went over some of the very good open source aspects of it, uh, of the previous version but this version still has, you know, it is still a vendor at the end of the day and so um, it doesn't happen all the open source aspects that a completely open source product might have like uh, N8N and things like that. Okay, um, that's it for this week. Uh, we'll continue next week going deep diving into the AI uh market and looking at AI tech broadly, how it's impacting small and medium sized businesses, uh, consultants and the enterprise. Hopefully this kind of overview of the most recent research on the top performing um m not legacy but more recently developed um, AI agent providers and AI agentic AI systems out there that are for sale. It gives you some idea of kind of what solutions you should be looking at for your business and what's currently available out there on the market for you these days. As always, thank you for joining us today. For more insights on AI governance and digital transformation please follow us and visit uh, cognicore.com.