Okay, Google: Can we talk about what I really want from an AI assistant?
Many of us have become familiar with the ability of generative AI to write some text for us in response to a prompt. It’s become so popular, in fact, that many tech companies are racing to add this capability to their products, often for an additional fee.
Also: AI is a lot like streaming. The extras add up quickly
A lot of value can be derived from creating text. It can cut down time, help focus ideas, help people who don’t write well produce professional-looking correspondence, and much more. ZDNET has been covering this aspect of generative AI all year. Gmail’s Help Me Write feature can compose a message based on a prompt, formalize it, extend it, shorten it, or try an entirely new draft.
But I want something more. I want Google’s AI to help me manage my email.
I have about half a million emails in my Gmail email store, dating back to the beginning of Gmail. I get hundreds of new messages daily. Gmail’s anti-spam filter does a fairly good job of blocking the most egregious spam from reaching my inbox, and I have a library of carefully curated filters that help me manage the rest.
Also: How to Write Better ChatGPT prompts to get the best AI results produced
But it’s still a daily effort, and it could be much better. Filters were introduced in 2004, shortly after Gmail launched. It hasn’t really been updated since then. We use approximately 20-year-old technology to manage the daily onslaught of message traffic.
I spend a lot of my day managing email. Of course, there are a lot of things I can’t delegate, like careful correspondence between team members and project partners. But I dedicate at least 30 minutes a day to managing flow.
Even with this level of time investment, my five main email categories—basic, promotions, social, updates, and forums—contained 41,330 messages. My approach is to let those messages that trickle in from the home page pile up, assuming that if there’s something really important, I’ll either track it elsewhere, or the person who sent the message will reach out to me again.
I want an AI assistant that I can rely on as an assistant
What I want is an AI assistant that I can train to manage all my current messages and all the messages that come in every day. Filters help, but they rot over time as email addresses and message headers change, and they don’t update to reflect new messages or moderated topics.
I want an AI assistant that I can collaborate with, and that can help me manage that flow.
I get a ton of press releases and pitches of all kinds from people hoping I’ll cover their products here on ZDNET. This is in addition to the regular promotions, newsletters, correspondence and other streams that non-journalists receive.
Also: 7 Advanced Tips for Fast Typing in ChatGPT You Need to Know
Over the years I’ve tried setting up filters for all the major PR companies, moving their messages from my basic category to my promotions category. I also drag and drop each press release from “basic” to promotions, which has also helped me with my Gmail training. But I now have 26,722 messages in promotions, including 114 messages that came in the last 24 hours.
To illustrate how this works, I will go through five examples. Keep in mind that the claims shown here are speculative. This feature does no Available in Gmail. Honestly, I hope the Gmail team will take this article as inspiration and add some of these features. If not, it may serve as an inspiration for other email vendors and those incorporating AI to think outside the generative box and consider how large language models can do more than just write basic prose for us.
However, let’s delve into the examples.
1. Sort press releases into Gmail categorization
I would like to give her some instructions. For example:
From now on, when a new press release comes in, remove it from your inbox and assign it to the “Press Releases” label.
Ideally, what you would do is: Find messages that represent press releases and save them in a file. This is actually better than filters, because sometimes I receive correspondence from PR companies that are not new press releases. They help me with a story I’m looking for, or I’ve found an interesting release that I’m following. I don’t want these letters in my press release folder. But if you filter by email address, for example, these details are not possible.
Also: Less typing, fewer errors: How Gmail snippets can save you time and effort
I also tried filtering by common press release terms, such as “for immediate release” or “ban,” but the filtering wasn’t global enough to be successful. AI is smarter than that, and should be able to separate press releases from correspondence, and know which emails represent press letters.
2. Unsubscribe from unread newsletters
Here’s another. I get a lot of newsletters, including a few I subscribe to. I read about five or six of them regularly. But most of it piles up, unread, because any interest I had that day has been replaced by the next thing I had to pay attention to.
So wouldn’t it be nice if I could tell Gmail:
From now on, unsubscribe from any newsletter I haven’t opened for 60 days.
Naturally, the AI must know what a newsletter is. Waiting 60 days gives me time to open something. But everything else can be removed, and this list will change over time.
3. Keep the latest promotional emails up to date
I get a lot of promotional mail from tool companies like Harbor Freight and Rockler. I also get a lot of photo and video companies. I appreciate these sale posts, because every now and then they have something I want. But it piles up. I currently have 962 items from Harbor Freight and 242 from Rockler, not to mention those from B&H and Adorama.
Also: Six skills you need to become a fast AI engineer
Again, filtering won’t help. And you won’t search. This is because – in addition to promotional information – I have purchase receipts and correspondence. I don’t want to delete those things.
With that in mind, wouldn’t that be nice?
From now on, when I get promotional mail from any of the photo or gadget companies, I keep the new one and delete the rest. Be careful to distinguish between promotional messages and purchase receipts. I keep all my purchase receipts forever.
4. Highlight emails from my teammates
I have a filter that looks for any email coming from zdnet.com and gives it a red ZDNET flag. This way, emails from ZDNET editors appear in my inbox. But this filter does not capture emails from other ZDNET contributors, many of whom do not have zdnet.com email addresses.
So, I’d like to be able to make this prompt to my AI assistant:
From now on, assign a ZDNET tag to every email that comes from someone with a zdnet.com email address. Every Monday, take a look at the team meetup page and get the names of everyone on the list. Assign a ZDNET tag to every email that comes from everyone on that list as well. Make sure to always assign these messages to the primary category and mark them as important.
5. Summarize the news that may interest me
So far we have looked at sorting and labeling. But we can use generative AI skills too. Wouldn’t it be nice to have a summary of the five or ten press releases that might be relevant to the work I’m doing?
Since it’s helpful anytime you create a claim, it’s a good idea to specify what you want the AI to do. So let’s do it here:
- Get tested every weekday at 8 a.m.
- Scan press releases and presentations that have appeared since your last scan. This will help cover releases coming out over the weekend. If we said “on the last day,” Monday’s scans would miss the stuff that comes on Friday and Saturday afternoon.
- Find presentations on a range of topics. For the purposes of this example, let’s assume that these topics are artificial intelligence, software development, and supply chain.
- Choose the five best presentations on each topic. We can define “best” as being from PR firms we have previously corresponded with, being complete (including vendor description and URL), and including at least one executive statement. There may be ways to improve these standards, but this is a good starting point.
- Determine how to present the information. I want the company name, a one-sentence summary of the press release, and a link to the original email.
Also: How to send password protected emails in Gmail
Okay, let’s try to turn that into a claim.
From now on, every weekday at 8 a.m., do the following:
Scan press releases and presentations that have appeared since your last scan. If this is your first screening, check press releases and proposals that have arrived within the last 24 hours.
Among the scanned press releases, select five each primarily related to artificial intelligence, software development, and supply chain. Choose only versions that include a resource description, URL, and at least one implementation statement. If there are more than five versions on each topic, choose the five versions that contain the most information.
Create an email message. Set the subject as “Press release summary for” and follow it with the date. Set the contents to be three bulleted lists, one for each topic. Before each menu, display the topic in H2 format. Each point will correspond to one specific press release. List the company name, provide a one-sentence summary of the press release, and then provide a link to the original email.
Send me this email.
This may take some fine-tuning, but it’s a solid idea. Instead of having to sift through all those emails, AI can prepare a daily file of the latest news.
Handrails and warnings
Some interesting challenges come to mind when creating a feature set like this, especially considering how error-prone AI tools are. If this capability is deployed, you will need a testing interface and an editing interface.
The test interface can be fairly simple. Gmail can simulate and then display different actions that AI might make to your emails, without actually doing anything. I recommend that this be in the form of snapshots in time, so that users can let the simulator run for a few days, see what would happen if the AI was unleashed, and then make changes to the prompts.
Also: The moment I realized ChatGPT Plus was a game changer for my business
This brings us to the second Handrail feature: the quick editor. Users will need to be able to go back and edit claims, especially when they are ongoing claims meant “from now on.” This can be a fairly simple interface, but optimization will be necessary with such open instructions and the range of possible changes.
I know I would greatly appreciate an AI assistant like this, but it’s important for both users and developers to proceed carefully. Google has a strong track record of releasing innovative features as betas, and I wouldn’t expect anything different here. As a user, it would be wise to adopt slowly, especially if there is no testing interface. Try very simple prompts, wait and see how things go for a week or so, and then add them to your instruction set.
Also: Google’s new tools help users verify photos online faster
Would you use this if it were available? Are there other capabilities you want? What will you ask your assistant to do? Let us know in the comments below.
You can follow my daily project updates on social media. Be sure to subscribe to my weekly update newsletter on Substack, and follow me on Twitter at @David GoertzAnd on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, and on YouTube at YouTube.com/DavidGewirtzTV.