Co-Founder, Chief Technology Officer of Slack & Director, Salesforce Inc
【本物のAIとゴミAIを見極めよ】セールスフォース共同創業者が警告「AIブームはITバブルと同じ道」／Slackは「最強AIツールになる」／AIを使えなきゃ「仕事はないと思え」【1on1 Tech】
... を破壊」の現実味 ◇出演◇ ▽パーカー・ハリス（Parker Harris） Salesforce 共同創業者／Slack CTO（最高技術責任者） 1999 ...Okay Parker Harris welcome to the show thank you so much pleasure to be here in Tokyo you're based in San Francisco san ...
About Parker Harris
At the TDX 2026 conference, Harris demonstrated updates to Slackbot, describing it as an "employee superagent" that can access company intelligence from conversations and documents. He emphasized the importance of user adoption, stating that "if no one's using it, who cares?" and noted that the company plans to focus on discovery and ease of use for all employees, not just engineers. During a Dreamforce 2025 fireside chat with Guy Raz, Harris discussed the impact of agentic AI, defining it as "an agent that can do things on your behalf" and predicting that a large majority of repetitive tasks will be automated. He argued that jobs will change rather than disappear, drawing parallels to the industrial revolution and the advent of personal computers, and expressed hope that AI could help reduce partisanship. Harris also reflected on Salesforce's founding, describing meeting Marc Benioff as "one of the luckiest moments in my life" and recalling the company's early emphasis on trust as a core value when customers were skeptical of internet transactions. He cautioned against corporations using AI as a reason to cut their workforce, stating that "you can't cut yourself to growth" and instead suggested retraining employees for higher-level tasks.
Transcript (41 segments)
I
Interviewer0:00
There's always a bubble when something big happens. We started in the dot-com bubble, and that's happening again. And I think with AI especially right now, people are going for the gold rush. How do you respond to investors' skepticism?
P
Parker Harris0:15
They want to discount that and say, 'Well, you know, AI is eating software.' And Slack is actually a jewel in our portfolio that is the best interface for AI. We're also hiring AI agents. People who no longer need to do the rote work, we could take those people and redeploy them. Please, you know, apply to these new roles. If you're not using AI in your job, you know, you should leave.
I
Interviewer0:51
Chapter.
Okay, Parker Harris, welcome to the show. Thank you so much. Pleasure to be here in Tokyo. You're based in San Francisco.
P
Parker Harris1:33
San Francisco. It's a long flight and jet lag. But, you know, I've been coming here pretty much ever since we started Salesforce Japan, which was in 2000. I started the company in 1999, so we're 26 years old now. I was trying to count how many times I've been to Japan. I think it must be at least 20. Which doesn't sound like a lot, but 20 times 20 trips is a lot of trips.
I
Interviewer1:59
Yeah, right. It is. Everyone, not only in Salesforce but also in the industry as a whole, is talking about AI, AI, AI and all things AI. The Gemini just launched a few days ago, and Nvidia had record earnings earlier today. How do you see the AI landscape right now from your point of view?
P
Parker Harris2:31
Well, first of all, I feel lucky to be in the middle of it. Salesforce's part of it is how do we bring enterprises into this new AI world with partners like Google, OpenAI, and Anthropic. I've never seen the pace of innovation faster. We have a research group writing papers on prompt engineering. I'd like to take credit for when OpenAI really hit that tipping point. We had a few ingredients in there, but we as a company were surprised. We had to repivot the entire company that year to focus on Agentic AI. I think every company has done that. What has changed in three years? The entire world has changed. You're talking about Nvidia. Who would have thought Nvidia, for gaming and graphics cards, would be where they are? I don't know who's going to catch up to them. OpenAI is an incredible company, but Google... On the flight here, I watched a documentary on the founding of DeepMind in London. It was very much an academic approach, not a business focus. Demis Hassabis... I can't pronounce his name very well, but it was a story of him and his history. What I loved about it was that it was about changing the world, not just about how to make money from it. I think this is a technology revolution like none I've ever seen. It has the power to do a lot of good and change a lot of lives, but there are also lots of risks.
I
Interviewer4:42
Right. It's the fastest pace of innovation since you guys founded the company.
P
Parker Harris4:55
I haven't seen much of it, but Salesforce started with the internet. The internet had been around a bit, but not for the enterprise. We started with a simple idea: don't buy your software. People were buying CD-ROMs, installing software, upgrading, buying servers, running their own data centers. This was 1999. We were born in that revolution. The next revolution was the phone. Now we expect to do everything from our phone. I order my groceries, get my dinner, everything in San Francisco is automated. I get my Waymo robot car with my phone. Why can't I do that in the enterprise? Each of these revolutions has been bigger than the last. Predictive AI was huge, but agentic AI is probably the biggest transformation we're going to go through in our generation.
I
Interviewer6:17
Right. But on the contrary, many analysts and fund managers are talking about an AI bubble. What's your take?
P
Parker Harris6:25
There's always a bubble when something big happens. When we started Salesforce, it was huge. Remember the dot-com bubble? We started in that bubble. I felt we had a simple business model: use the internet to deliver Salesforce automation in a better way. Then there were all these other dot-coms asking, 'What's your business model?' It was all about eyeballs and impressions. Some of those companies became massive, but there was a wipeout of companies that didn't make it. That's happening again. Anytime there's a revolution, people go for the gold rush, and not every mine has a lot of gold. With AI, you can create an incredible demonstration or startup, but it's harder than it looks to make AI successful. There will be a lot of failures, just like the dot-com bubble. But you'll also see the next new giant companies emerge. Salesforce has the potential to be one of those, to ride this and bring our customers into the agentic revolution.
I
Interviewer8:15
Do you feel like you and everyone at Salesforce are inside the gold rush, or is everyone doing the gold rush thing at Salesforce right now?
P
Parker Harris8:29
Yes, everyone is. We have 70 to 80,000 employees worldwide. We had already written our business plan for the year, but when OpenAI hit that tipping point, we took a step back and replanned. It's like a battleship; we don't turn in a moment, but we can turn. We realigned our priorities to say AI is the most important thing. Every single person at Salesforce is focused on AI. We're saying everything needs to be AI-first, Agentforce-first, and Slack-first. Slack is a jewel in our portfolio, the best interface for AI. Everyone is part of this revolution. If they're not, that's a problem. We'd tell them to wake up. If you're not rushing forward, you're going to get left behind.
I
Interviewer9:51
Right. You said the organization is going faster now, but Salesforce paused hiring engineers at the beginning of this year?
P
Parker Harris10:05
No, those are crazy rumors. We are probably growing less at the net level in headcount, but we're still hiring. People leave, we have attrition, and we're hiring the best people. We're also hiring AI agents. Our support organization looks very different today. We have support people who no longer need to do the rote work, like password changes. The agents can do that. So we can take those people and redeploy them. We need more forward-deployed engineers. We told people to apply to these new roles. This is the upleveling of human capabilities through AI. If AI is going to do the rote work, people will find new jobs. They might need retraining or more education. That's what's happening now with every big technical revolution.
I
Interviewer11:38
Everyone's talking about AI taking jobs away. Anthropic's CEO Dario Amodei says AI is wiping away entry-level jobs. What's your take?
P
Parker Harris11:53
Jobs are going to change dramatically. Do you need a paralegal as you think of them today? Or an investment banking research analyst out of college doing research? They're probably not going to do the same thing. If you're not using AI in your job, you should leave. If you're not being trained to use AI in education, you're not going to get those jobs. Jobs will change dramatically, but I think there will be more jobs. If we have more capabilities, we'll dream up more things to do as a species. All those opportunities will need more people leveraging these tools. Some AI agents will do work for us, some will do work with us. It's still going to be humans and AI, and we're going to be elevated as a species because of it.
I
Interviewer13:03
You're also CTO of Slack. Salesforce acquired Slack in 2021. Everyone uses business messaging software. You call Slack an 'agentic OS.' How does it work as an OS in this era?
P
Parker Harris13:26
You have to roll back to the vision of Slack's founding, which I take no credit for. The vision was that it is the center point to do work. Slack is most successful in engineering groups. They use it because it's multiplayer, with multiple engineers working together, connected to their tools. All the different systems are integrated into Slack. Now with agentic AI, Slack is that central point connected to your agents. Agents come into Slack directly to you in a direct message, or in a channel where work gets done. Slack is a place where work gets done, not just a place to send messages. The agents are now in Slack with the humans. We have a feature called Slackbot. We now have a new Slackbot that is unbelievable. We're using it internally and with a few pilot customers. It could be one of the best features, and it's going to be that hub where work gets done.
I
Interviewer15:08
What kind of platform will Slack be in the future?
P
Parker Harris15:14
Slack will continue to be the place where work gets done. Any collaboration tool has a lot coming at you. It's a lot of messaging. People say it's sometimes distracting. We take that problem on head-on. How are we going to do that? We're going to use AI to prioritize what's important. We'll have agents handling some of that, clearing it out. We'll clean up the interface with AI. If we get it right, the UI will be more dynamic. You'll be living in it, with channels, messaging, humans, and AI, but you'll be looking at what's most important each day. It'll tell you what's happening. It would be great to have my agent tell me, 'Here's your calendar, here's what's important, here's the briefing for what we're doing right now.' I have this on my calendar, but there are documents, structured data. Pulling that all together, running it through LLMs and Agentforce, gives better answers. It's a problem, like Gmail with email coming at you all day. Many companies are going after this, but we have the user base. People love Slack, and I want them to keep loving it. I'm going to do everything I can to keep making it better, easier to use, with less noise and more intelligence. The magic of Slack is not inventing the future and knowing what it is. It's working with the users and inventing in micro-increments, knowing when we have product-market fit.
I
Interviewer17:34
When it comes to enterprise agents, context is very important. Microsoft and Google have productivity software suites like Google Workspace and Microsoft 365 with calendar, mail, documents, and spreadsheets. They have more context about how workers work every day. Do you think Slack and Salesforce software can compete with them?
P
Parker Harris18:17
I don't see it as a competition because we're bringing all that context into Slack. We're connecting to Google Drive, calendar, Office 365, depending on what tools you use. Context is king. We're bringing those assets in. There's a lot of context in Slack and in Salesforce. We've been thinking about context for a long time. We are a data company at our heart. Now, instead of diving into the depth of all that data, we have the AI dive into it and tell us what we need to know.
I
Interviewer19:05
As a consumer, everyone is using ChatGPT, Gemini, or other AI services. But an MIT report said that 95% of companies fail to use AI to generate returns despite billions in investment. Why is that happening in business environments?
P
Parker Harris19:39
We're all picking up ChatGPT, Claude, Perplexity, or Gemini to answer questions and learn. You're using an AI model with data baked into it. It's a fairly simple construct. But in the enterprise, the data needs to be correct. ChatGPT didn't train on every corporation's data. Enterprise data is often locked up. Everyone had anxiety about being left behind, so every company said, 'I want AI in every department.' They gave everyone a license for an LLM and said, 'Go forth and do AI.' People created use cases that looked great initially but weren't repeatable and had no security model. They stumbled. It was the fear of missing out, the gold rush. That's where we come in. We've built a platform to help customers have an integrated platform where applications and data are connected and unlocked across the enterprise. We give our customers Salesforce automation, marketing automation, e-commerce, and they make it their own. We're doing the same for AI. We have agents out of the box, but every company is different. With Agentforce, you can build your topics, skills, connect it to your data, and iterate.
I
Interviewer22:40
Are AI agents good enough to be our coworkers?
P
Parker Harris22:47
In some cases, they can be, but it's super early. We have a role at Salesforce called the Sales Development Rep. We hire young people to call customers and generate leads. We now have an SDR agent built on Agentforce that goes after leads in our database. The agent pulls a lead, sends an email, and if the customer replies, it's a two-way conversation. The agent can have a multi-turn conversation and pass it on to a human to close, or occasionally close the deal itself. That's a coworker. We have an agent on our help site handling two million conversations. Williams Sonoma has a cooking concierge. We're seeing it in many places, but it's not everywhere yet. A corporation might have armies of autonomous agents doing lower-level roles so humans can do higher-level work. That's a grand vision, but we're not there yet. Anyone who says agents are taking over the enterprise is completely wrong.
I
Interviewer26:00
Agentforce adoption is not high yet. You have 12,000 adoptions out of 150,000 customers, and about 6,000 are paid. Where do you see adoption going?
P
Parker Harris26:22
Adoption is definitely exponential. Customers need to get their data right. Many are ahead on that, cleaning up data sources and unlocking siloed data. We have forward-deployed engineers going into customers and iterating with them. Everyone is at a different point in the journey. Customer agents on websites answering questions is one of the first use cases. We've proven that out. No more useless chatbots. Companies are happier and more productive. Customers are happier getting faster answers. From there, we go into sales, marketing, and employee productivity. It's an incremental move across all the touchpoints of the enterprise. It's a race. We are well-positioned, but we need to go fast. We have 12,000 now, and we need to scale that.
I
Interviewer28:01
People used to say 'software is eating the world.' Now some say 'AI is eating software.' What's your take?
P
Parker Harris28:20
Both are true somewhat. AI is software. If Salesforce stood still and didn't change, AI would eat Salesforce. Someone would do it better, just like the internet ate enterprise software. Salesforce rode that. Everyone uses SaaS now. You're just part of the next revolution. I don't think software is dead. You can't just... if I were a model company, I'd say we're putting billions into data centers to build massive models that will do everything. But I don't think that's true. You need structured software. But will you use it the same way? Will you log into Salesforce as an enterprise application? Or will AI decompose enterprise software? That's more my belief. Decompose means I get the pieces I need in the moment I need them. If I'm in Slack talking to AI, the AI brings me the functionality I need in the moment. The user interface is no longer one giant interface; it's dynamic. We have all the breadth of Salesforce, but we're reconfiguring it into this new world that's AI-first and Slack-first.
I
Interviewer30:34
So that's why you call Slack an agentic OS.
P
Parker Harris30:34
Exactly.
I
Interviewer30:42
But investors seem to be worried about where sales are going in the AI age. How do you respond to their skepticism?
P
Parker Harris30:48
I tell the customer stories. Some people, because we are successful, want to discount that and say AI is eating software and SaaS is dead. Those people's companies have not been as successful in enterprise software or CRM. They don't have those customer stories, so they tell false stories. We tell customer stories, like at the event here in Tokyo. People doubted us when we started Salesforce. They said no one would trust their customer list on the internet. We believed in it. With every revolution, there are doubters. I believe in our vision and success. We will beat everyone there.
I
Interviewer32:11
You said SaaS is not dead, but everyone can code and develop software with AI now. Everyone can build their own workflows. Maybe eventually everyone can build their own CRM. Do you think Salesforce will survive in the AI age?
P
Parker Harris32:43
Absolutely. AI can write code, and it's great at it, but also terrible at it. We've seen huge productivity gains, but it's really when engineers know if the AI is doing a good job. Entry-level programmers using AI can create 'AI slop'—software that isn't great and creates productivity loss. You could tell AI to code a UI that looks like Salesforce and give it a database, but do you have the taste for what good software is? Are you iterating with it? How are you going to support it? You have a demonstration. We will use AI to go faster at Salesforce. We're already 30% more productive in engineering because of AI. If we're going faster with what we have, we're racing into the future. You might use AI to code something like Salesforce was 10 years ago, but we're moving so fast you won't catch up. You can create a UI quickly, but what's underneath? Do you have a deeply integrated platform of applications, data, and integrations? Do you have a conversational interface, deep analytics, and industry-specific capabilities? There's a lot there.
I
Interviewer35:33
Earlier you talked about the documentary on DeepMind's CEO Demis Hassabis. People in Silicon Valley talk a lot about AGI. When we have AGI in the future, what would the enterprise or software look like?
P
Parker Harris36:20
I still think we're a ways off from AGI. Even the best models hallucinate on simple questions. We saw an example where it put one of the founders of Twitter as the CEO of a hot new AI startup. That's not true. If that's still happening, we're not that close to AGI. If we do get to AGI, the idea of what humans and agents will do together will be transformational. What will Salesforce look like? We have to iterate our way there. It will be a different company. It's an exciting future for civilization, but we're a ways off. We may hit another AI winter. We've trained on all the data we can and are creating synthetic data, but are we really going to get to AGI quickly? We have amazing tools now that companies like Salesforce can wrap around LLMs with platforms like Agentforce and make them useful. That's the innovation we'll see in the near future. The models will keep getting better, but I don't think we're at AGI anytime soon. If I were a model company, I'd argue we're getting to AGI because I'd be asking for more capital to build data centers and buy more NVIDIA chips. But I don't think it will take a while. At Salesforce, we don't need to debate it. We have amazing tools today. Our challenge is to make our customers successful with them. We're not waiting for AGI.
I
Interviewer39:07
Great. It was great to hear all of this. This was a lot of fun. Great questions. Thank you.
P
Parker Harris39:07
Thank you. Thank you for coming.