#tbd - Even the bot leaves your website: why bad search is now twice as expensive
AI is radically changing search in e-commerce: users expect guidance instead of hit lists, AI agents shop on their behalf, and companies control their search directly from AI tools via MCP server. In the new insights! episode, Christina Schönfeld, Director DACH at Algolia, explains why the SaaS apocalypse is not happening, which software vendors will nevertheless lose out in the agentic world, and why hundreds of A/B tests per week will soon be the benchmark. This article summarises the key takeaways and gives you concrete impulses for your AI and commerce strategy.
Joubin RahimiManaging Partner · synaigy7 min. read
“The speed at which the market is moving right now is something a great many companies underestimate, because you simply no longer have the time to evaluate for a year, then implement for a year, and then take a look. You need that iterative approach. In German culture you are engineering-driven and you plan a lot, because if you are building a bridge, it had better be well planned. Culturally, Americans just try things out, and if it does not work, they change their minds. That is exactly what we are seeing right now in the AI world - how important that is.”
Your keyword search has had its day: how customers really search today
Do you remember the last time you typed just a single keyword into a product search? That is exactly the point. Christina Schönfeld, Director DACH at Algolia, has spent years watching search behaviour in e-commerce shift: long-tail queries are increasing massively, meaning rare, specific queries that regularly send classic search systems down a dead end. Users describe their problem in full sentences, rephrase it, try variants, and still expect the right result immediately. At the same time, attention spans keep shrinking. Anyone who fails to display the right product at breakneck speed, ideally personalised right on the landing page, has already lost.
The technical challenge behind this is bigger than it sounds. Your search still has to respond accurately when someone simply types T-shirt, with or without a hyphen. But it also has to understand what is meant when someone asks for a shirt for a Pilates workout. Being able to do both at once, precision and contextual understanding, is the real core of modern AI search. And when search does not deliver right away, it should hand the user on elegantly: to a chatbot, for example, that uses a RAG system to access reliable product data instead of hallucinating, and makes fitting suggestions based on the description. Search thus evolves from a search box into guidance.
Agentic commerce: when your most important customer is a bot
Perhaps the most important insight from the conversation: it has long since stopped being only humans who search on your website. AI agents such as agentic browsers or tools like Claude Cowork visit shops, click, filter and search on behalf of their users. Christina puts it wonderfully dryly: if the search does not work, the bot loses interest in staying on the site too. That sounds amusing, but it is hard business reality, because the agent that gives up simply takes its human's purchase order to the competition.
For you, that means your digital infrastructure will have to serve two target groups in future, humans and machines. Specifically, you should check:
Can an agent find the right product on your website within seconds, or does it fail because of poor search and an unclear structure?
Is your product data clean, complete and prepared to be machine-readable?
Alongside classic search, do you also offer conversational access such as a chatbot with a RAG connection?
Is your platform built API-first so that agentic systems can interact with it at all?
Anyone answering these questions with no today is optimising past a customer journey that is currently changing fundamentally.
Hundreds of A/B tests per week: the agentic backend is coming
While everyone is talking about changing user behaviour, Christina's heart beats faster for a different development: the way companies will work with their search in future. The old world, in which teams laboriously maintain synonyms and click their way through dashboards, should definitely be history. Instead, teams pull their analytics directly into AI tools such as Claude or ChatGPT via MCP server, have optimisation potential pointed out to them, and then send off an agent that sets up the campaign and launches the matching A/B test along with it. By its own account, Algolia was the first vendor on the market with an MCP server and is currently learning at high speed which use cases customers are implementing with it.
There is one figure from the conversation you should remember: it cannot be that companies run only one A/B test per week or per month, it should be hundreds per week. That is not humanly achievable, but it is with agents. The vision behind it: you define your strategic goals and work together with agents to steer your company, instead of turning every single dial yourself. Just as important as the automation is control. In an enterprise environment in particular, you have to define precisely which agent may access which indices and data, and who is allowed to change what. After all, the working student is not allowed to do everything either, so why should the bot be?
SaaS apocalypse? Why open APIs decide your survival
If agents do the shopping in future and developers can rebuild entire software products with AI, will specialised vendors still be needed at all? Christina answers this deliberately pointed question with refreshing composure. She sees the SaaS apocalypse invoked in so many podcasts in a far more nuanced light: the value of a solution lies not only in its features, but in years of learning around a problem. That is exactly what many in the market vastly underestimate. For her, the current debate is above all one thing: build vs. buy on steroids, an age-old discussion with a new turbocharger. She explicitly welcomes the fact that the commoditisation of development work suddenly enables everyone to implement their ideas, because the difference is then made by creativity and strategy.
Her real warning is aimed at a different group: software vendors that are not open. Anyone who does not think API-first, does not maintain good documentation and has never invested in their developer community will have a hard time, because agentic systems simply cannot work with such black boxes. And her advice to you as a decision-maker is delightfully concrete: if a vendor promises you that you do not have to worry about a thing and should simply call them, ticket and white-glove service included, then be sceptical about whether that model still holds up in the agentic world. Openness is no longer a nicety for developers, it is becoming a criterion for survival.
AI first instead of mobile first: 40 percent faster in the project
Because Christina promptly turns the tables and asks questions herself, this episode also offers a rare look into synaigy's engine room. Joubin reports on a project at an industrial company that was set up entirely AI first: AI in development, AI in analysis, AI in coordination. The result speaks for itself: the project was around 40 percent faster than originally planned, with less effort. It was not yet profitable, but as a foundation and a learning curve it was priceless. The decisive success factor was less the technology than the mindset: allow experimentation, catch concerns centrally so the team can simply carry on without negative vibes.
That experience yields a principle that goes far beyond individual projects: speed is the hardest currency in the market. Anyone who traditionally submits tenders two minutes before the deadline is giving away a strong signal. Anyone who delivers significantly earlier simply shows: we can do this. With AI, you can read in specification documents, run estimates and produce proposals in record time; the north star at synaigy is to put out a proposal within 24 hours. And if customers are already writing their RFPs with AI while service providers answer them with AI, then we humans have long been the middleware between two agents. All the more reason to keep reviewing the results critically.
Conclusion: iterating beats perfecting
If you take a single message away from this episode, let it be this: the time for years of evaluation and waterfall projects is over. Christina sums up the cultural difference perfectly: German, engineering-driven thoroughness is great when you are building bridges, but in the AI world the winners are those who work iteratively, try things out and adjust course as soon as new insights arrive. Figure it out as you go is not chaos, it is the appropriate answer to a market that changes on a weekly cycle.
Whether it is AI search, agentic commerce, MCP servers or A/B testing by the hundreds: all the topics in this episode share a common denominator, namely the courage to start early and learn along the way. Watch the full insights! episode with Christina Schönfeld and pick up the impulses for your own digital and commerce strategy. And if you want to know how an AI-first project could work for you too, just get in touch.
Would you like to listen to the full episode?
Here you will find the podcast of the interview:
Would you rather watch the episode? No problem!
Here you will find a recording of the interview:
Please accept functional cookies to watch this video.
Would you rather read?
Joubin Rahimi
Fantastic that you are joining us again for a new episode of Insights. My name is Joubin, Joubin Rahimi, and joining me today is Christina Schönfeld from Algolia. Hi Christina!
Christina Schönfeld
Hello!
Joubin Rahimi
Nice that the two of us get to talk again. We were just puzzling over it, it has been two and a half years.
Christina Schönfeld
Yes, unbelievable.
Joubin Rahimi
Exactly, time flies. And I know that after the podcast we had quite a few conversations with people who were immediately interested, and they then took a look at Algolia, because you were one of the first, if not the first company to have AI in search. Well, you still have it in search, but I think you were the first.
Christina Schönfeld
Yes, indeed. Yes, we implemented it very quickly, in various elements of course.
Joubin Rahimi
That was even before ChatGPT, before they got all the hype.
Christina Schönfeld
Well, AI has been around for quite a while, and we were applying machine learning even before that, actually. Of course, generative AI added a completely new element, and we were really quick off the mark there. But we were also the first solution to offer an MCP server at all on the market, so we implemented that very, very quickly. And yes, we like speed, it is part of our DNA after all.
Joubin Rahimi
That is great. I have already dived far too deep into the questions. Would you like to say a couple of words about yourself and about Algolia, for everyone who did not see the last episode and does not know Algolia?
Christina Schönfeld
Two and a half years ago, yes. So you can already tell, I have been with Algolia for almost 8 years now. Back then I opened up the DACH market, closed the first customers and then also built up the team. And yes, it is going very well, still, it never gets boring. And yes, we have grown very strongly and were able to win a lot of new customers, especially in the past year.
Joubin Rahimi
And yes, I find it so nice, Algolia, a European solution, and in that respect
Christina Schönfeld
Oh yes, that really is an important element, because our founders come from France, so we have European roots, and that too is written deeply into our DNA, how we handle data, data security, yes. And at the same time an incredibly international company, with headquarters in the USA, in San Francisco still, and an absolutely international culture.
Joubin Rahimi
If you have been there for 8 years, that speaks for itself.
Christina Schönfeld
Oh right, yes.
Joubin Rahimi
For the company and for you, for both. AI, you were the first, or machine learning, to build it in, and is that actually true?
Christina Schönfeld
I do not know whether I can sign off on that. I would first want to check it journalistically, whether we really were the very first.
Joubin Rahimi
Okay, then we can do a fact check afterwards. In any case, for us, from what I know, you were very early on when it came to machine learning and the like.
Christina Schönfeld
That is true.
Joubin Rahimi
Now AI has really been hyped to death over the past 2 years, and it has changed a lot too. How has search changed for you? Has user behaviour changed? Have requirements changed?
Christina Schönfeld
Yes, I think we are seeing 2 strands, as I always call them, of change. Of course the behaviour of end users on a website or in an app. And then on the other side, that is the second one, how companies actually want to work with a search. And to address the first one, user expectations: we already noticed, I mean, right before the ChatGPT launch, that these long-tail queries were increasing, and by that I mean queries that are not necessarily long words or many words, but ones that occur rarely. And that is not possible-
Joubin Rahimi
On the platforms where I work, on Google.
Christina Schönfeld
Exactly, so that was already strongly on the rise before generative AI came along, and now of course it is even more the case, because users simply expect to be able to rephrase their project and still find the right information. And we all know our attention span keeps sinking and sinking and sinking. And with the sheer volume of information that is of course even more pronounced. So you have to display the right product at breakneck speed, and not just in the search results but really when you land on the landing page, ideally personalised, no matter who arrives, to the person, so recognising the intent very quickly. And what I find interesting is that even bots play a role. It is no longer only humans coming to the website, but also the bots, which search in just the same way. And I do not know whether you have tried it and how it is now, but when I tried out Comet, the browser from Perplexity, a few months ago just as it had launched, and wanted to visit some retailer. Wow, the amount of clicking and searching and filtering that went on, and it really took a long time. And now with Claude Cowork too, if the search does not work, then the bot loses interest in staying on the site as well.
Joubin Rahimi
That is a nice one too, even the bot cannot be bothered to stay on the site any more.
Christina Schönfeld
Yes, or you just abandon it, because of what you want to do. Yes, so those are new challenges. And then of course you move away from just navigating or typing in a keyword search, and instead you want guidance. Because a great many companies are thinking about how they can have a chatbot too, so that users have a reason to come back to their site. And within a chatbot like that with the LLM, because LLMs love to hallucinate, as we all know, you need a RAG system, that is, a retrieval engine, and that is us again. And yes, there is a lot you can try out there. I know it myself, I am currently renovating an apartment and I look around the marketplaces, DIY marketplaces, and I cannot work out everything I need. Instead of laboriously clicking around and searching, I really often go into the chatbot and ask what I need, and then I get my suggestions there.
Joubin Rahimi
Okay, so that is changed user behaviour in itself. So basically you enter sentences more and more, or users tend to enter long sentences. Is that what you meant, or rather many terms?
Christina Schönfeld
Both, one as much as the other. And that is exactly the challenge, it is so varied, because you still need the accuracy, so that when someone enters T-shirt, with or without a hyphen, you quickly find the most relevant thing for you as a user. But then you also have T-shirt for, I do not know, for my workout, my Pilates workout, in there as well, or things like that. And technically speaking it is really not that easy to link those together.
Joubin Rahimi
To understand both, right?
Christina Schönfeld
Yes, at the same time. And when you do not have the... So when you have a longer input or do not find the right thing directly, you make suggestions within the search auto-suggest already. Hey, I cannot find that right now, but by the way here is our chatbot, and then you describe more, it works out the context from the description and suggests results again.
Joubin Rahimi
The second thing you mentioned was the backend, so how companies, what companies want from search, or is that what makes your heart beat faster?
Christina Schönfeld
That is what makes my heart beat faster, because I find what is happening there so exciting. And I think a great many people are of course talking about changed usage behaviour and agentic commerce. It is even more exciting when, I think, companies now, you know this too, have an incredible amount on their roadmap, have an incredible number of ideas they want to implement, and mostly it fails because of some bottleneck within the organisation. Exactly, and I mean, this old world of search, well, there is always something you can tinker with. So, and in the old world people set up all sorts of synonyms, that should definitely no longer be a thing. But where we actually want to get to is that companies can not only automate but also keep control, of course, we are always about control, that is everything to us, and can set up tests. Lots and lots of A/B tests. And someone once said, maybe it was even our CEO, it really stuck in my head, it cannot be that our customers set up only one A/B test per week or once a month. I mean, they should be running hundreds per week, and of course that is no longer humanly possible. But it then moves toward planning in MCP servers in various capacities. Yes, please, give me the data here, the analytics from Algolia. That is already possible today. And then you get that played back, and those are only the beginnings. And then the next thing you say is, okay, now help me set up a campaign, or the analytics, exactly, what should I optimise? And then the agent runs off, optimises it, and then of course you set up the A/B test along with it. And you look and keep optimising. And I think ultimately the vision goes toward you having, alongside your strategic goals, your focus point, and then working together with the agents to steer your company.
Joubin Rahimi
So an example, let me try to take an example. You have a B2C example now, but you can probably translate it to B2B as well. You have a stationery business, quite a big one, and you sell everything. And all of a sudden one particular item takes off, Legami pens. The kids love these Legami pens, they write well and so on. Personal experience, let us put it that way, they always want something from Legami. Yes, and then the search results do not have to be right at the top at all, but they are somewhere with a strong trend, and that is what you can filter out well. So with your approach, I pull out the analytics, I pull out the AdWords, analyse where I have trend changes, in order to recognise, hey, the brand Legami seems to be going through the roof right now in their-
Christina Schönfeld
Yes, yes, exactly, and of course that should happen automatically, or it already happens automatically today based on click behaviour. So that the ranking changes. And of course, if you, you know, you like buying the pens for your kids regularly and have a preference.
Joubin Rahimi
Daddy, daddy, daddy, daddy!
Christina Schönfeld
What else could be relevant for you as a user?
Joubin Rahimi
I have to be honest, they get bought straight from Thalia.
Christina Schönfeld
Okay.
Joubin Rahimi
Thalia does a good job on both sides there. But that is another story.
Christina Schönfeld
That is a completely different story.
Joubin Rahimi
The backend itself, you said you have an MCP server. What else can you do with it? So what else do customers use it for?
Christina Schönfeld
Okay, we are learning that ourselves right now and finding out. And what I know from our product managers is that we are really going in to look at what the actual use cases are that they want to implement. And I would have to ask again what all came out of it, because for us too the releases right now are changing the statistics so, so quickly. That is really exciting. So what I definitely know is that you are in Claude or in ChatGPT and you ask, hey, give me the analytics for the last few days, weeks and so on, and that is the classic case. What will come next is that you will also be able to change things within the dashboard. We will probably start, presumably, with your synonyms and so on. But right now it is a long time frame. What is also interesting is that you really have to think about it, especially in the enterprise space, what user rights you give the bot, the agent. And that is the first thing we solved. So we gave that direct thought.
Joubin Rahimi
May it search all products or not, or which ones?
Christina Schönfeld
Exactly, which index, but you know that, that is also the case at Algolia, that there are lots of configuration options. That is something companies do not necessarily think about during evaluation, but it is very, very important that you check, okay, can I restrict my products and indices, can I say that my working student may only configure this and that, or nothing at all, and so on.
Joubin Rahimi
That is often the difference between enterprise systems and non-enterprise ones. On the feature map in a tender like that, everyone can always do everything, but the question is how.
Christina Schönfeld
Yes, yes, exactly, everything in detail.
Joubin Rahimi
Exactly, the devil is in the detail.
Christina Schönfeld
Exactly, the devil. They have been configuring that for a long, long time, otherwise we would not have grown the way we have grown.
Joubin Rahimi
With regard to Algolia and search, now that agentic commerce is here, you do not need an interface at all any more. Let me overstate it. Then we do not need search at all any more. If someone came up to you here at the trade fair like that, what would you say?
Christina Schönfeld
That really is an exciting question, is it not? Because ultimately I have listened to a lot of other podcasts, of course about this SaaS apocalypse that is taking place, and how suddenly everyone can do everything. I see it a bit more nuanced, because it is about more than just what features you have, it is also about all the learning around it. And I think a great many people in the market right now underestimate how much software solutions and vendors have engaged with certain problems. And that is what I believe. That is why we exist. I found that very exciting just now. And then this morning, I heard this, now, because with Codex and Claude Cowork too there is a discussion about how interesting it is that these companies, both companies, are pursuing such different strategies. Because purely in UX a strategic decision was made by OpenAI, we believe everything should be the same and uniform, and Anthropic says: no, we want it, we believe there is value in keeping it separate. And that also means that on this topic UX plays a huge role again, and you have no idea what end-user behaviour will necessarily be.
Joubin Rahimi
I think we are at the beginning and nobody knows exactly where it is heading.
Christina Schönfeld
But what I do believe in any case is that software vendors that are not open, that are not API-first, you know, that do not have their documentation as well, that have not invested in their developer community, will find it hard, because agentic systems will then have less information available to interact with those systems. That is my theory. And yes, I think you should give it a lot of thought when a vendor promises you: hey, we solve everything for you, do not worry about a thing, just call us, ticket and white-glove service. I do not know whether that will still work in the agentic world.
Joubin Rahimi
For one thing, I find it exciting that you said we have then disclosed everything. We have another customer who had their software rebuilt from the documentation using AI. So the other way round. Reverse engineering. Yes, the question really is how much AI systems can suddenly reverse-develop from existing software. But I think that is always only half the truth, because then you may have a piece of software that can do it. You have to operate it, maybe you then have Base44 operate it or something. Exactly. But you still always need people, because things change. And they keep taking care of it.
Christina Schönfeld
I think, for me it all sounds a bit like build vs. buy on steroids. This discussion has been around for ages. And yes, I think that is completely fine.
Joubin Rahimi
But now there is a moving target for developers. Yes. They do everything themselves.
Christina Schönfeld
I actually think, not just the developers, I think it is really great that everyone is now being empowered and enabled to implement and try out their ideas. That is the commoditisation of development work. And then it comes down to, okay, what really sets me apart as a company? And that is the creativity behind it and how you think about your strategy.
Joubin Rahimi
Or in your case too, how you arrive at the results, right? That is it, and what options you have to influence them.
Christina Schönfeld
Yes, exactly, control again, exactly, yes, yes, indeed, that you do not give everything out of your hands and no longer know what is happening, but instead work with the AI in that moment, yes.
Joubin Rahimi
So, before I ask you another question, in the pre-talk Christina said, can I ask you questions too?
Christina Schönfeld
Yes, of course. Yes, exactly, right, good. I am actually always the one asking the questions, this is very unfamiliar for me. So how do you see it? Because how do you perceive it when you talk to your customers? Do they, yes, when they hear AI, do they tend to think, oh, that is still a long way off, or do they go all in?
Joubin Rahimi
After they have talked to us, they no longer think it is a long way off. Fine, but before that. Most of them are trying something out and are at least at that level. Whether management and the budget owners are already doing so is another matter. At the latest when we come in with proposals where they see that we definitely work with AI and also make assumptions or have requirements, when we say this is what we would like, then you can have the price discussion, and then you notice, okay, how much are you doing with AI? And then a conversation about it comes about. Because we do a lot, we develop with AI, we have products that are based on AI. With customers themselves it depends heavily on their maturity level, but many, especially in the lines of business, quickly use something, because you can do that privately too, right? ChatGPT for 20 euros, even 100 with Claude, and suddenly you see that even if you have not connected anything, you can still upload things and you notice what an advantage it gives you. Yes, and at the lower levels, so the people working operationally, there are some of these front runners who simply try it out. And the question is, what kind of ground does this company have? Can these people do more or not? If yes, then it moves forward. From our point of view, few companies have a plan out in front and drive it through centrally. But I also believe that driving everything through centrally is not the right thing, because you simply leave the creativity on the table.
Christina Schönfeld
Yes, and do you see any commonalities between the companies, so are there specific levers where you would say, ah, that is why this works well, so fewer processes? Or who actually steers processes like that? Do you see, are there Chief AI Officers walking around anywhere?
Joubin Rahimi
I really have not seen that title anywhere. But I think it is the wrong one. Everyone has to deal with this new technology. Chief AI to me always means going into the executive board, but really you have to move the managing director there. We now have, and this is unrehearsed, we really work with another company, Ecodynamics, which prepares training like that for managing directors for Handelsblatt and FAZ, and we underpin it with additional content, just for those, a one-day training so that the person simply understands what this is, when he looks at business, who has already done such things. So that those people understand it, because it really does change a great many things fundamentally. And I think that is the commonality. Those who have not started with it tend to be a company where the freedom is not there, either because it is not lived or is not there, and where leadership is not pushing it either. Yes, and yes, we at least want to enable those who say, I think there is something there, and then they only have the management and afterwards you exchange ideas. So I think that is the lever there. The structures and the freedom have to be created from the top down, but a lot is also already coming from the bottom up and people just do it.
Christina Schönfeld
Yes, of course people do. Well, that is the exciting thing, that right now we are all in a joint science-fair project, where everyone is experimenting and that then takes on a dynamic of its own. Of course you always have to think about it, especially when you are working with customer data, that you are super careful, and that is why I also understand many companies that want to take a step back first.
Joubin Rahimi
But the exciting thing is of course also the customer data, there is other data worth protecting too. Now we are at our end, we have a project at an industrial company, Schmersal, I think that is a shared customer of ours. We have been live since Monday too, so very fresh, and there we said we are going in on the AI side, AI first, so not mobile first, but AI first. Our initial proposal was simply way over the customer's budget. They also had someone who was under budget, but as I said, I do not trust that. Everyone was over except that one, and I would fundamentally like to do it with you, but you are over my budget. And then it took us a good half a year until we came together. But we only came together by saying we do everything with AI, not everything is weighed on the gold scales, even though we have specification and requirements documents, so very clearly. We went in classically and we are then allowed to move the data into the AIs, and that does not mean we cannot leave this or that out, and the project was so much faster than we had initially planned. We were 40% quicker, the effort was lower. We learned a lot, which is why it was not a profitable project, but it was the foundation for everything that follows. So, why did I tell you all this? Because everything they did was constantly trying out new things, sometimes with ChatGPT, with GitHub, Claude, we want to connect this, or connect that, and then the data protection people at our company raised concerns. We always said, whenever something comes up, bring it to me, whenever that comes up, to me. I wanted them simply not to have to argue and not to have any negative vibes, but to be able to just carry on. And of course we do have data worth protecting. We have defense companies, we have public services, we know the private addresses of many people, which you should not have, so not all of us, but in context. Of course that must not be touched in any way, that is completely clear, but most companies have Microsoft in use anyway, Google, whatever, and push the data into a company subject to the FISA Act anyway, and then it is in the USA anyway. So sometimes the discussion is also considerably overblown. Okay, yes, I welcome everything that is in Europe. Now I am talking far too much. Christina has revved it up.
Christina Schönfeld
That is how it goes.
Joubin Rahimi
I understand that, then she is in a good mood.
Christina Schönfeld
Exactly. Well, now I have forgotten my question.
Joubin Rahimi
No problem. We can do a wrap-up. That is what we are seeing, and the software developers who also took a critical look at it, with the project, those who were in the project said: okay, that fits. Is the code perfect every time? No, but people also apply double standards.
Christina Schönfeld
I remember, I think we talked about it back then, when we last spoke, or maybe, if I am not mistaken. Because I think what is so fascinating, what you are describing, is that it really is a totally iterative project, so figure it out as you go. And I think, because you just mentioned you needed half a year to even get to the close, so in evaluation probably even longer, and the project then ran-
Joubin Rahimi
Close, just the sales close, where actually everything was clear and you could have signed. That was maybe three quarters of a year.
Christina Schönfeld
Yes, and I think the speed that is happening in the market right now is underestimated by a great many companies, because you simply no longer have the time to evaluate for a year and then implement for a year and then take a look, instead you have to have this iterative approach. And I think, back then that is why, podcast, I have been saying this for many years, our dynamic in German culture, you are engineering-driven, you plan a lot, because if you build a bridge it had better be well planned and not collapse. And the Americans, that is cultural, they just try things out, and if it then does not work, they change their minds and then it starts again. Exactly, and I think that is what we are currently realising in the AI world, how important that is.
Joubin Rahimi
Speed is a lot. Now the customer could have been faster, and so could we, of course. What we have taken away from the project is, or when did we win deals? When there is a tender and it says the deadline is in 3 weeks on Friday at 12 noon, when did you typically submit? Friday at 11:58 in 3 weeks. Very, very often. We changed that a few years ago. To submit as quickly as possible. Of course you always give up the opportunity on price. So if you have picked up something you were not supposed to pick up and so on, or can give it more thought, if you delay it you still have the option to change something. If you submit early, you cannot change it. But what we have found is that if you submit early, really significantly earlier, it simply shows you can do it. And with AI we have now read in specification documents like that and run the estimate. Because you can, software developers can estimate endlessly, and the longer they estimate, the better it does not get. Yes, it only ever gets more expensive. And then, okay, that has to go into the AI, and then a contract comes out. So we are not there yet, but that is the basic idea, where we say when a customer comes and describes things however they describe them, I want us to be able to put out a proposal within 24 hours. Simply as a north star.
Christina Schönfeld
Yes, that goes straight on to my team here. We are professionals after all.
Joubin Rahimi
Because when the customer says: okay, wow! And that then fits with all the details the customer supplied, then that is great.
Christina Schönfeld
I also find it very amusing, because you increasingly notice that customers of course use AI to write their RFPs. And ultimately you then have, we are basically, we humans are the middleware. So really it is two agents side by side.
Joubin Rahimi
That is completely fine too. And then we have this huge thing, and when you as a human have to work through it, you also think, wow, can I really change the logo in the top left? So, you really do get a tender like that, 3000 lines, the logo must be replaceable. Okay, you want a headless - what is this rubbish question doing in there? We did that with AI. All great. But if you can then also answer it with AI, it does not bother you at all any more.
Christina Schönfeld
Yes, but you do have to read it over. So we always find that it is very, very important that you still check everything again, and I think we are not quite there yet with the AI systems.
Joubin Rahimi
Yes, that is a good thing too. A customer said, you have 6 weeks. You send it and a day later he says, here is the proposal, we can start. And then I say, yes okay, now we wait another 6 weeks. Because you can pamper the customer for another 6 weeks, great, and you know we can do this, we can do that.
Christina Schönfeld
Okay, so, whoever is listening to this podcast right now and is not yet in it, feel free to test it or send in your RFP. I like it even better when nobody sends me an RFP at all, but just has a conversation and then the evaluation starts, of course. But yes, we are happy to accept RFPs and can open up a challenge.
Joubin Rahimi
Christina has hijacked everything now. Thanks for watching, and if you have questions about Algolia, search, search in general, drop them below or via LinkedIn. I think you can be reached there too.
Christina Schönfeld
Yes.
Joubin Rahimi
We did not agree on this at all, but who is not on LinkedIn these days.
Christina Schönfeld
That is it.
Joubin Rahimi
It was a pleasure.
Christina Schönfeld
Thank you very much.
Joubin Rahimi
Thank you.
Do you have questions or feedback?
Then feel free to contact us directly.
- Joubin Rahimi
Managing Partnersynaigy
Show phone numberShow mobile numberShow email address
Subscribe to the blog now and never miss any news
✔️free ✔️news every week ✔️expert knowledge
Please accept the corresponding cookies to view this embedded content.