Is chat the right interface for AI-native B2B software?

Geoff Charles · Dec 6, 2023

AI will make software look far less like tables, graphs, forms and dropdowns and much more like a dynamic, conversational interface that abstracts the data and surfaces the intelligence directly

Customers' jobs to be done can be enumerated cleanly and are few, so AI can just do them — as with Ramp replacing manual review of a transactions table with automatic flagging of out-of-policy or fraudulent charges

41:22 20VC: Ramp's Product Playbook: How To Hire Product Teams, How to Run Sprints, How to Increase Product Velocity, When and How to Go Multi-Product with Geoff Charles, VP Product @ Ramp

Jesse Zhang · Sep 19, 2025

The exciting part of the space is not rising resolution rates but the expanding scope of the AI agent — a good agent becomes a conversational UI and concierge for the brand, engaging users pre-purchase, driving revenue, and acting proactively rather than reactively

Solving inbound support inquiries is table stakes; senior leaders are excited because the agent can become a new brand interface that generates revenue, so success metrics shift to retention and conversion rates

Scope: will become 'a lot more real' in three to five years

23:52 20VC: Why 90% of Founders Build Startups Wrong | Why AI Growth Rates are Sustainable & Remote Work is BS and the AI Talent War | Competing with Brett Taylor and Sierra: Who Wins the Customer Service War with Jesse Zhang, Decagon

Jesse Zhang · Sep 19, 2025

Moving from support to a general conversational brand interface requires one more level of abstraction — designing any conversation, including open-ended outbound ones where the agent must persuade rather than execute a procedure

Support has a fixed dynamic — the user comes inbound and you run a procedure — whereas open-ended conversation (e.g. charming a user into activating an account) breaks that framework

Scope: Decagon is doing some of this with customers already but it is 'still pretty early stages'

26:42 20VC: Why 90% of Founders Build Startups Wrong | Why AI Growth Rates are Sustainable & Remote Work is BS and the AI Talent War | Competing with Brett Taylor and Sierra: Who Wins the Customer Service War with Jesse Zhang, Decagon

Clay Bavor · Jul 4, 2026

It is inevitable that companies will interact with their customers through sophisticated agents that capture everything the company knows and can do, with conversation as the interface, rather than pointing customers to websites.

Agents can handle the complexity and get the job done on the customer's behalf.

43:51 20VC: Open Models vs Frontier Models: Who Actually Wins? | The $100,000 Token Budget Every Engineer Will Need | Why Forward-Deployed Engineers Are the Future of Enterprise AI with Clay Bavor, Co-Founder of Sierra

Ishan Mukherjee · Apr 4, 2025

There is a new UI paradigm for AI software sitting between a chat interface and traditional flat SaaS, which is more task-oriented

Rox looks like a CRM but is actually a swarm of agents working for the user

38:03 20Sales: How the Best Sales Teams Use AI to Win Enterprise Deals | Sales Teams Will Be Dramatically Smaller | How to Ramps Sales Reps Way Faster | Why Unpaid Design Partners are BS | Why this Generation of Sales is Soft with Ishan Mukherjee @ Rox

Max Junestrand · Jan 26, 2026

Legora's first product in summer 2023 was the wrong direction because it was built around click-and-point core use cases with no agent or chat operating over those tasks

37:21 20VC: From Only OpenAI to Die-Hard Anthropic: The Downfall of OpenAI in Enterprise | Harvey vs Legora: Legal AI is a Winner Take All | $7M ARR in a Single Day and Raising $200M Across 3 Rounds with No Deck with Max Junestrand, CEO @ Legora

Becca Lindquist · May 2, 2026

Clay's Sculptor solves the blank-page problem by letting a user describe the desired research and outreach in natural language and having the table built for them

You can state a prospecting goal conversationally and Sculptor constructs the table, so you no longer start from nothing

59:47 20VC: Inside Clay's Sales Playbook Scaling to $100M ARR | How to Set Sales Comp Plans | How to Read Sales Talent Linkedin Profiles | What Profiles to Hire & Fire | How to Increase Performance and Speed in Sales Teams with Becca Lindquist

Christian Lanng · Sep 27, 2023

Natural language alone is not enough — prompting won't replace interfaces, and a genuinely new generation of user interfaces is yet to be invented

Everyone said you can do anything with ChatGPT but 99% of people use it as Wikipedia; mobile similarly took five years before responsive design concepts emerged

48:49 20VC: "How Being a Founder Almost Killed Me"; We Have Lied to a Generation of Founders | The Hardest Truths About Being a Founder Revealed | Why AI Co-Pilot is BS, Seat Pricing is Over & User Interfaces are Stupid with Christian Lanng

Cem Kansu · Jun 20, 2025 · speculative

Current 'AI-first' products are just text or voice boxes; the real unlock is AI-generated UI created on the fly rather than hard-coded interfaces

On-the-fly UI could follow wherever a learner's conversation and interest leads — e.g. generating a mountain-hiking scene to teach the word for mountain — making the experience far more entertaining

Scope: twenty-year horizon

73:41 20Product: How Duolingo Build Product 10x Faster with AI | Duolingo's Biggest Lessons on Paywalls, Push Notifications and In-App Purchases | Why Small Teams are the Future of Product | Why PMs Will Become Extinct with Cem Kansu, CPO @ Duolingo

Aidan Gomez · Aug 19, 2024

Chat is the right interface for some tasks but not as the interface onto everything; traditional GUIs remain the better choice for much software

He doesn't want to type out explicit instructions for everything — sometimes clicking through a GUI is the faster way to get the job done

Scope: chat is still a compelling interface for some uses

20:38 20VC: Chips, Models or Applications; Where is the Value in AI | Is Compute the Answer to All Model Performance Questions | Why Open AI Shelved AGI & Is There Any Value in Models with OpenAI Price Dumping with Aidan, Gomez, Co-Founder @ Cohere

Josh Browder · May 18, 2026

Chat is the right interface only where the job is converting unstructured user input into structured choices; it is a bad interface for other consumer AI use cases such as travel

With parking tickets, consumers had no idea which of the twelve valid cancellation reasons to pick, so letting them write anything and matching it to the correct defense was the unlock; travel doesn't have that structure problem

Scope: good in some areas, terrible in others; travel named as the bad case

74:45 20VC: Turning Peter Thiel's $100K into $10M Angel Portfolio | The One Man Accelerator at The Four Seasons | Why VCs Can Be Sharks and What Founders Need to Know | Why Stocks and Cash are BS and You Should Invest in Land with Josh Browder

Ishan Mukherjee · Apr 4, 2025

Chat interfaces are powerful but too limiting for daily work in B2B software, because end users expect the software to do the work and lay out its outputs as workflows rather than answer queries

B2B end users come in to get a job done and want pre-created outputs like account reports, org charts and queued email tasks rather than asking the system for things

38:50 Chat too limiting b2b users want prepared workflows

Mike Cannon-Brookes · Oct 13, 2025 · hedged

The interface for AI will be evolutionary and end up looking much more like today's software than people expect, just smarter and with a much lower click-to-value ratio

Users want stability — in user studies they rely on knowing which button does what — so constantly AI-regenerated interfaces would be too discombobulating

24:35 Interface evolves gradually staying familiar not regenerated per screen

Anish Acharya · Feb 9, 2026 · hedged

Voice is amazing for enterprise, but dynamic UIs and chat UIs are overstated in consumer; browse-based interfaces will largely stay the same

Chat is the optimal UI for the most high-agency people who can articulate exactly what they want, but many consumers are looking to spend time rather than save it, cannot always articulate what they want, and prefer to browse

23:40 Chat fits high agency users consumers still want to browse

Daniel Khachab · Oct 28, 2024

Product design shifts from drawing rich user interfaces to designing the character and personality of the AI agent the user talks to

If a multi-step workflow like payroll collapses into a single prompt, there is no UI left to design — what remains is deciding who the user is talking to and how that agent behaves

7:52 Product design shifts from ui to agent character and personality

Daniel Khachab · Oct 28, 2024

SaaS is dead and will be replaced by the agentification of software

User interfaces exist only because software has to be learned; when you prompt a machine in natural language there is nothing to learn, so agent-based tools will be adopted faster

10:53 Agentification replaces saas since natural language prompting removes the learning curve

Max Junestrand · Jan 26, 2026

The Claude Code / Cowork paradigm — give an agent an overarching task, have it return a plan, then let it roam across tools and MCP servers to execute — is the right direction and should be applied to legal work

It mirrors how a partner works with a senior associate and a senior associate works with an associate, effectively adding another layer to the hierarchy with AI at the bottom, available to everyone 24/7

19:26 Agent as junior associate task then plan then tool roaming

Phil Libin · Jan 4, 2017

The move to conversational interfaces is a major disruption, not an evolution, because it breaks the one-person-one-technology paradigm into multi-human conversations that bots participate in

For decades all technology interaction has been one person poking at one system; with bots the technology becomes present in a conversation among multiple humans, which is hard to imagine as anything but completely disruptive

14:44 Multi human conversation breaks the single user paradigm

Alexander Embiricos · Feb 21, 2026 · hedged

Productizing prompts and human actions is OpenAI's job, but the most value right now sits in an intermediate stage rather than in the fully productized magic-text-box end state

It's hard to know which specific AI feature or form factor for a specific market will actually work, even though many companies are attempting it

10:45 Open ended intermediate tools carry more value than productized form factors today

Alexander Embiricos · Feb 21, 2026

The next phase this year is a shift from text boxes that return information to text boxes where you state a task and something actually gets done for you

Google search was the box for navigation, ChatGPT the box for information; tasks are the natural next step, even if the agent only does part of the task

31:11 Text box shifts from returning information to executing tasks

Alexander Embiricos · Feb 21, 2026

Conversational interfaces (chat or voice) will be the enduring pillar of AI interaction — a single entity you can talk to about anything and add to any group chat — but paired with bespoke graphical interfaces for power users doing specific work.

You shouldn't have to navigate to separate places for a coding AI versus a sales AI; but being able only to talk to a capable assistant is as annoying as an executive assistant you can never work alongside — at some point you want to open the artifact and edit it yourself.

32:32 Chat as universal pillar with bespoke guis for power users