On the agenda: artificial intelligence and internet search. Microsoft has supercharged its Bing search engine by teaming up with OpenAI, the startup behind the buzzy ChatGPT, to create what Microsoft CEO Satya Nadella described as a “new paradigm” in search. 

Instead of traditional search results, Microsoft’s new A.I.-powered Bing provides “answers” to questions. And Bing’s newfound chat prowess means you can converse with it, asking follow-up questions with the same natural language you might use when talking to a friend. 

While A.I.-powered Bing is still not available to the general public, Microsoft gave attendees at the event hands-on demos, along with a series of briefings from company executives discussing some of the finer points.  

Here are a few first impressions and key takeaways from Fortune’s experience at the event: 

1. The line between ‘holy crap’ moments and gimmicky moments is still not as bright as it needs to be to replace search 

Because ChatGPT has been available for a couple of months, many people have already experienced the mind-blowing moment of seeing generative A.I. technology in action for the first time. Seeing the tech in a web search context is very cool and impressive, but perhaps not as astonishing as it would be if it came completely out of the blue. 

Some of the impressive moments can be subtle: I asked Bing for a three-day travel itinerary to New Orleans and it spit out a list of sights and activities spread out over three days. Was this just a list of things to do or was it a viable agenda that I could follow each day and have enough time to accomplish the various suggestions? I pressed Bing for more info in the chat section, requesting a chronological daily schedule, with travel time between activities. It delivered, showing me exactly how all the suggested items fit together chronologically. 

At other times though, the A.I. answers and chat can feel a bit gimmicky. When asked for Ethiopian restaurant recommendations in New Orleans, the A.I. churned out two paragraphs explaining basic facts about Ethiopian cuisine—interesting, but not really what I was looking for. Finally, in a third paragraph it listed two Ethiopian restaurants in New Orleans, along with menu highlights and hours. Was this really any faster or more useful than just typing “Ethiopian restaurants New Orleans” in Google?  

2. It’s way too early to call a winner in this race

The dueling press conferences and CEO blog posts from Microsoft and Google this week were intended to deliver a clear message: This is a big deal. But they also had a strange odor of hastiness to them, as if the two companies were racing against each other in order to be first. What’s not clear is who is reacting to who in this race. 

Google has clearly been caught off guard by ChatGPT’s popularity and reportedly went on an internal “code red” alert to speed up the release of its own conversational A.I. technology, which it has been working on for years. Google now says its Bard technology is available to “trusted testers,” with more access opening up in the “coming weeks.” But it remains to be seen how widely available it will be, or how the product will compare with the Microsoft-OpenAI offering. 

Microsoft gets kudos as the first out of the gate with a real, usable product (as opposed to just a canned demo). But people’s search habits won’t change overnight, and right now it’s too early to tell how big of a lead, if any, either company actually has. 

3. Microsoft wants you to know that it is being very responsible when it comes to A.I. 

The main keynote event included a presentation by a Microsoft executive overseeing Responsible AI, and a session later in the day was devoted entirely to the subject, with Microsoft president Brad Smith participating in the panel. The Microsoft executives spoke at length about the many layers of oversight and safeguards they have implemented around artificial intelligence, stressing a commitment to transparency, without providing all that much detail.  

The exercise seemed primarily intended to blunt any unflattering contrasts with Google, which has made a big deal about how it’s moved slowly and cautiously because of the inherent dangers involved with A.I. And coming seven years after Microsoft’s fiasco releasing a racist and foulmouthed A.I. chatbot named Tay, the company knows the importance of establishing trust with the public and policymakers ahead of any potential blunders its latest A.I. might commit in the future. 

4. Microsoft does not want investors to know the price tag of this A.I. adventure 

The company repeatedly sidestepped questions from reporters and analysts that tried to get an idea of how much more expensive it is to serve up A.I. search answers versus traditional search results. OpenAI CEO Sam Altman has previously said that the cost of each ChatGPT transaction is in the single-digit cents—a level that gets expensive when hundreds of millions, or even billions, of people are using a service, and which, according to a recent Bloomberg report, is well above the cost of a traditional web search query.  

Microsoft exec Yusuf Mehdi demurred when asked about the cost by Fortune’s Jessica Mathews during the keynote Q&A, noting that, in addition to being confidential, the costs would be tough to calculate because each Bing search is different—some rely on the new A.I. tech to answer a question while others rely on traditional search engine algorithms. 

Microsoft chief financial officer Amy Hood was equally opaque, noting in a separate briefing that costs will come down over time, but that even now, each new user or usage of A.I.-powered Bing represents “incremental gross margin” for Microsoft.