My (Mostly) Unfiltered Thoughts on the New Vision of Tableau

Reflections from the 2025 Tableau Conference

This past week, I attended the Agentforce Tableau Conference. It was a big week for Tableau / Salesforce, because they have a new vision / product, and this was a big opportunity for them to sell us on it.

This is going to be a longer post, but I wanted to take the time to type up my thoughts, reactions, and key takeaways / next steps from this conference.

The Conference

I came into this conference with an open mind, but I will admit this to you - I had not been very impressed with anything I had seen about Agentforce / the New Tableau so far. I wanted to come out of the conference with a better sense for the vision of where Tableau will go in the future, where I fit into it, and what it says about BI in general.

At the conference I talked to a few people at the various Tableau Next / Agentforce booths, attended the Keynote, and also got the chance to sit in on a personalized demo / conversation with our Tableau pre/post-sales team for a discussion about our specific use case (at my company). I got a very different feeling from our personalized session than I did anywhere else, and so I want to capture the full picture in this post.

I have not done a deep dive of research into making sure I am completely understanding each product in the new lineup, nor the complete vision. If you are looking for a more definitive source for your product-purchasing decisions, I would recommend the Tableau documentation, or videos from Tableau Tim or the Sensemakers.

What follows is my interpretation and reaction, and I will not be doing much additional research on the products as I write this up :)

What They Are Selling

The bold new vision of Salesforce/ Tableau is agentic analytics, aka an ‘AI agent’ that takes care of everything for you from data prep, to visualization creation, to taking action based on surfaced insights.

The way this works with the current product lineup relies on one main SKU (stock-keeping unit… aka the thing you buy) - Tableau+. Here is a link to information about the full lineup of products. At the bottom of this article you can find a writeup of the key products that are included in Tableau+, if you’re interested.

I’m going to start with the punchline and tell you what my takeaway from everything is, and then I will unpack what I think this means for the future of BI, my impressions of the products/marketing, and finally a breakdown of the actual products.

The Next Steps I Personally Want To Take

I do not make purchasing decisions for my company, so I can’t tell you what our strategy or next steps will be. What I can tell you is my personal reaction, and what decisions I would make if I was working within my own closed-system.

I think that Agentforce is likely worth purchasing right now - I could see it speeding up and improving a lot of sales tasks right away, without any iterations to the product needed.

I do not think I would purchase Tableau+ at this time. It seems like it is still being built, and I do not want myself / my users to be the guinea pigs who are discovering the sticking points. I also feel that my team is pretty good at the ‘Last Mile’ of analytics (building dashboards), and don’t see any of these products solving the ‘First Mile’ (pipelines, modeling, etc). More on this later.

While it would be cool to have an agent that could help people build dashboards more quickly, and I am sure people would use it, I am not convinced it is worth purchasing for.

However, it is something I would be willing to put aside time to do a trial of, and potentially setting up a willing pilot group of users to test out the possibilities. I enjoy giving product feedback, and I think that it could be fun to test out all of the new features, and write-up my thoughts. Earlier this year I tested out Tableau Pulse, and put together a many-page document with the things I did and did not like about the product.

Apart from Tableau/Salesforce, my main next steps are:

• Continue on my quest to centralize all important business logic in our database (not the viz layer!) so that our database is a Single Source of Truth which is ready for good analytics or AI - using any platform we want.

• Explore semantic layer options and improvements - from public handbooks, to internal documents, and better column naming to new data catalogs, I think there are many places that we can be preparing ourselves for broad AI use by improving our semantic layer.

• Test out different options! I would like to try out a Tableau+ trial, while also continuing to test things like Snowflake’s Cortex AI features, and Claude Enterprise.

My Impression of Marketing and Keynotes vs. Personalized Use Cases

Coming into the Tableau Conference I was very skeptical of the entire Agentforce / Tableau Next topic, mainly because I think the marketing has not really been resonating with me. There are so many names for all of the different products, and the names change often! It has felt like Salesforce is assuming people are excited about Salesforce, so they are seeing all of these different products from Salesforce and getting excited by them.

I just find it confusing - we have Tableau Next, Tableau Plus, Tableau Pulse, (Salesforce) Data Cloud, Tableau Semantics…. And more. Especially for the Data Cloud product, I think myself and many others were worried that we would need to pay to store our data on Salesforce Data Cloud in order to use the new products, and there was confusion if we could only do this with Salesforce Data or not.

In the keynote / booths I could tell that Salesforce gave everyone clear instructions to try to dispel myths/ clarify confusion - yes you can use Tableau Next with non-Salesforce data, yes Data Cloud has zero-copy abilities so you are not storing 2 copies of your data, yes your Tableau Cloud data sources are available to you in Tableau Next.

What Happened to the Analyst?

The other sticking point I had coming in was something that was pointed out by the Action Analytics podcast in this video (timestamp: 16:42) - Tableau’s marketing used to put the analyst at the center. It was all about the analyst getting into their flow. Now, Agentforce / Agentic Analytics is the center of their marketing… and it makes the analyst feel left behind.

I have not felt like I could see myself (an analyst) in any of their marketing, which is confusing to me, because who are the people setting up, and enabling Agentforce? Your analysts. As someone who not only builds dashboards but also works with stakeholders regularly, I know that we are still essential to the process - even if we have an agent that helps speed up building dashboards.

I genuinely wonder if the people who wrote the keynote watched the video I linked above, because it seemed to demonstrate a somewhat defensive response to say “here is how the analyst is still involved, we have not forgotten about the analyst!!!”

In the demonstration, the CPO of Tableau role-played being an analyst, and setting up an analysis workflow for the CMO using Tableau Next / Agents. There was a dialogue between the two of them that showed an analyst curating a data source, and walking a CMO through using it and then driving insights.

I appreciated the time they spent on showcasing the analyst - in a demo at the Tableau Conference. This felt like an important thing for them to do, and I can imagine they would not have used precious keynote time on a demo like this if it was Dreamforce (the Salesforce Conference) instead of the Tableau Conference.

First Mile Analytics are Still Unsolved

The last impression / hesitation I had from the marketing and keynote was something else Jonathan mentioned in the video from before - ‘Last Mile vs First Mile’ analytics. When I look at the capabilities of Tableau Next (building visualizations, answering questions based on metrics), I see a lot of ‘Last Mile’ actions. Although these can take time, I feel that generally my team has a pretty good handle on the ‘Last Mile’ items.

The thing that we all spend the most time on is ‘First Mile’ analytics. You can’t just build a dashboard or an agent off of your raw, messy data. First you need to clean it, model it, and shape it so it is ready for analysis. Agentforce doesn’t seem to really be tackling this.

I did not come away with much of a changed mind on this point - I think there is some calculation assistance that the Agent can help with, but I still think that if I were to buy Tableau Next tomorrow, my team’s energy and brainpower would still primarily be dedicated to First Mile analytics, with little assistance from Agents.

Why Not Just Sell Agentforce?

When my team had our personalized session with our Tableau pre/post-sales team, my focus was actually not on Tableau at all - it was on Salesforce. We got to ask questions about how we could use Agents right in Salesforce to do things like monitor opportunities and send alerts when they reach a certain threshold of number of days without stage progression / activity.

It makes sense to me that Salesforce, a huge company, would build a product (Agentforce) which works really well for their customer base - Salesforce users! Coming out of that session, I could see many possibilities for how Agentforce could be a truly great tool in the toolbelt of our Salesforce users - without even involving all of the ‘Source of Truth Analytics’ we are doing in Tableau.

Honestly, I wonder if it is even in the best interests of Salesforce to ‘care’ that much about the future of Tableau. They kind of have it made right now - they have the codebase of Tableau, and have been able to integrate that right into their already successful platform. Using Agentforce, Salesforce users can get visualizations and analytics support in the pages they visit every day, with a few extra clicks over to the “Tableau Next” interface (in Salesforce) if needed.

Long-time Tableau users who have a love for Tableau Desktop, Tableau Cloud, and Tableau Public (or a need for Tableau Server) would hate to hear this analysis, because we would like to continue to see innovations for the platforms we use, without having to switch to switch to [Salesforce] Tableau Next.

I heard a lot of fear and apprehension at the conference from Tableau lovers around a fear in the possible future divestment of Salesforce development power in any innovations for the Tableau that many people know and love. In the keynote Salesforce was very clear that Tableau Cloud/Server is not going anywhere, and that they are still investing in it, but it seems many people are still skeptical how long this will last for.

I personally hold the opinion that Salesforce has built a great Salesforce product, and so as disappointing as it would be to see Salesforce ‘kill’ Tableau, I could imagine it might just make more sense for them as a company that operates under capitalism.

How This Ties to a Wider Vision for BI

Despite my skepticism around Tableau Next / Salesforce branding, I came out of this week with a clearer understanding of the broad vision. I think that the ostentatious marketing that Salesforce has been participating in, combined with the ten million names for different products, has been a distraction from the real vision that Salesforce is trying to sell.

I do agree, generally, with the direction that Salesforce has taken, and I am curious to explore how companies could follow this path - with or without Salesforce. I think the general vision can be boiled down to a three-part system:

1. A database(s) where all of your raw / modeled data lives.

2. A semantic layer with rich information about your data. Ideally, this is agnostic, and can connect to and from any source you want.

   1. This includes documentation, further modeling, and other context.

3. A ‘report’ layer which is what your users interact with. This can be a variety of outputs such as:

   1. Your BI tool (Tableau, Looker, Power BI)

   2. An LLM (ex. Claude, ChatGPT, Agentforce)

The Future of BI - Includes a Semantic Layer that can power BI or agents/AI

I feel that Salesforce has put their brand so heavily on the semantic and report Layers that it clouded the vision for me, but now I get it. If you could get a setup like this it could be incredibly powerful, especially if everything within was vendor agnostic. The power that could come from connecting any LLM to not only data, but a well formed semantic layer, would change what it looks like to be an analyst. For example, Claude can generate beautiful and informative dashboards with just a few sentences of input, and it takes no technical skill to ask it to refine and change any visuals you are unhappy with.

Edited to add: I don't think this is entirely different from what many companies are already doing. Many of us use a database, have some kind of semantic layer (or two or more), and use a reporting layer (or two or more).

The difference is the addition of LLMs, which calls for a more robust semantic layer.

Salesforce is trying to be the one platform that does it all, no need to bring your own semantic or report layer.

But Salesforce is not the only option. I think the reality we will see become possible is this:

• A robust semantic layer that is agnostic in both directions - towards the database, and towards the reporting.

• A semantic layer agnostic multi-faceted reporting layer.

And some companies will want to purchase these capabilities as a package (Tableau+), and others will want to choose their own adventure.

In high tech I see choose your own adventure as more probable, because highly technical teams like the freedom to choose every piece of the puzzle, and build their own stack where possible.

Products Included in +

Tableau Next

An analytics platform like Tableau Cloud, except built right within Salesforce. Your Tableau data sources from Tableau Cloud are available for you to use in Tableau Next, but your users would need to use two separate logins. It seems like the dev team is still working on all of the ways that they can make the integration between these two environments as seamless as possible.

Agentforce / Tableau Agent

I can’t actually tell if Agentforce is included with Tableau+ or only ‘Tableau Agent’, but for this article I am going to treat Agentforce as part of what you get when you buy Tableau+

Agentforce is AI agents that are powering your analytics experience. You can use Agents to do a variety of things - help you create calculated fields, chat with your Tableau Pulse metrics, kick off a wide variety of actions within Salesforce, send Slack messages based on insights from your data, and summarize metadata for better semantics (basically documentation, if ‘semantics’ feels like a foreign concept to you).

I am only really scratching the surface of the things Agentforce can do, I am pretty sure there are people whose entire jobs right now are explaining all of the many things Agentforce can do, but this is a good little summary.

Metrics / Tableau Pulse

It seems to me like Tableau has taken the idea of ‘Tableau Pulse’ that they have been marketing so hard for the last year, and integrated it into Tableau Next. But instead of calling it “Tableau Pulse Metrics’ and capitalizing on that (albeit limited) momentum, they have just called ‘Metrics’ - and left ‘Tableau Pulse’ to be a feature that is part of Tableau Cloud - not Tableau Next. Are you confused with all of the names yet?

I might be wrong about this, but in the demos I saw, it seemed like Pulse was treated as a separate thing than the ‘metrics’ of Tableau Next.

If you are not familiar with Tableau Pulse metrics, they are essentially KPIs. You connect to a data model, and select a ‘metric’ - like Revenue. Then, you choose which dimensions from your data model you want your users to be able to slice it by (time dimensions, geography, product name, referral source, etc), and your metric is born.

Then, users can subscribe to this metric and get regular updates in Slack/ their email when it changes. They can also ‘chat’ with the metric (in a very basic chatbot kind of way) to use natural language instead of relying on dropdowns.

Data Cloud

In order to really leverage Tableau Next and Agentforce, you need to use Data Cloud. I am still not clear enough on what Data Cloud is to be able to explain it here, but it seems to just be a place where Salesforce can access your data from.

I heard a lot about open access and zero-copy in Data Cloud, which seems to be intended to avoid storing your data in two places. I know that Tableau Cloud is integrated with Snowflake in this ‘zero copy’ way - so Tableau Next and Agentforce can look at your Snowflake data without any of it needing to be ‘stored’ in Data Cloud.

Tableau Semantics

This is described as “Centralize your data and metrics in a single, governed layer for seamless analysis and efficiency”. I did not see this in action very much, but I get the premise. Tableau semantics is your semantic layer.

The semantic layer contains documentation, but also allows you to do things like flexibly form relationships between published data sources. It also contains ‘semantic learning’ - so it can learn from your Q&A with agentforce and store that knowledge so that your agents are more informed next time you talk to them.

---

Did you attend the Tableau Conference? Are you following their messaging? I am interested to hear how others are seeing themselves in this ‘Future of BI’ that we all seem to be racing, yet crawling, towards.