NEA’s Tiffany Luck says enterprises are still figuring out their AI ROI | Equity Podcast

Tiffany Luck's Background and Career

• Tiffany Luck, a partner at NEA, has been in the industry for a while and has been focusing on early-stage AI, APIs, and B2B SaaS, and she joined NEA at the beginning of 2023 10s.
• Before joining NEA, Tiffany was a partner at GGB Capital, and prior to that, she worked on the operating side, doing customer acquisition, business development, and marketing, including being one of the early employees at a NEA portfolio company called Lot 18, a wine flash sales startup 2m6s.
• Tiffany also spent time at Amazon in the early days of Amazon New York, working on consumer packaged goods, and she had to convince CPG manufacturers that e-commerce was the future, which has since become a reality 4m42s.
• Tiffany's experience also includes working at Morgan Stanley in the tech investment banking group, where she worked on M&A and IPOs, and she is now being tapped for her expertise on the consumer bit and the future of personal agents 6m15s.

The Concept of 'Magic Moments' in AI and Personal Agents

• The concept of "magic moments" in consumer technology is being discussed, and Tiffany thinks that personal agents are almost ready to become a thing, and she loves the way the phrase "magic moment" is used to describe significant developments in the field 8m40s.
• The conversation is expected to dive deeper into the future of personal agents, this year's AI IPOs, and how startups are helping enterprises track return on AI spend, with Tiffany's expertise being leveraged to explore these topics 10m10s.
• Creating a "magic moment" with AI, whether for consumer or enterprise, is a key goal, with examples including a first-time ride in a Whimo, which is a highly personalized and exciting experience, and AI-assisted healthcare, which can provide second opinions and help with medical decisions 10s.
• In the enterprise setting, AI can help with tasks such as putting together work projects and facilitating hard conversations, with tools like Claude being particularly useful for these purposes 2m6s.
• On the consumer side, AI is being used to develop personal assistants, such as Town and Ali, which aim to manage the user's mental load by taking care of tasks, managing calendars, and remembering important events, with the goal of creating a seamless and trustworthy experience 4m42s.
• The development of personal assistants is still in its early stages, and current AI models, such as Claude and Chat GPT, can be used as assistants, but they often require teaching and can create more friction, rather than simplifying tasks 6m15s.
• The ultimate goal for AI-powered personal assistants is to create a "magic moment" where the user can trust the assistant to take care of tasks and manage their mental load, without needing to double-check everything, and companies like Town and Ali are working towards achieving this goal 8m30s.

Privacy, Security, and Trust in AI Personal Agents

• The main concerns that hold people back from using personal agents are privacy, safety, and security, as they do not trust that these tools will not create more areas for hacking, 10s
• The idea of having an AI that can see everything a person is doing, such as logging into their bank account, is still a step too far for many, but there are a lot of smart people trying to figure out how to address these security concerns, 2m6s
• Companies are emerging that specialize in AI-specific security, and there are also new certifications and standards being developed, such as the AI underwriting company AIU, which is working to create standards for AI safety, 4m6s
• The development of personal security agents, such as a "bodyguard" or "agent guard", could be a potential solution to address security concerns and provide an additional layer of protection for users, 6m6s

Foundation Companies and the AI Ecosystem

• The conversation also touches on the role of foundation companies, such as OpenAI, which has filed confidentially for IPO, and how they can impact the development of AI technology, with OpenAI focusing on the consumer use case and Anthropic focusing on the enterprise use case, 8m6s
• The current time is seen as exciting for public markets, with companies like SpaceX, Anthropic, and OpenAI being considered game changers that have created real business value and grown quickly, with the market and ecosystem expected to benefit from their growth 10s.
• The lines between enterprise and consumer businesses are expected to continue blurring, with companies like OpenAI and Anthropic having use cases in both areas, and their growth is anticipated to continue, with the market considering factors like compute cost and sustainability 42s.
• Investors will be looking at unit economics and other factors to determine whether companies like Anthropic and OpenAI can achieve significant growth, such as becoming trillion-dollar companies, and to assess their potential for free cash flow and meeting public market expectations 2m6s.
• The ability to access compute, owning models, and owning infrastructure are all considered valuable, with the view that AI is creating value at all layers of the stack, including the infrastructure, model, and application layers, and that this rising tide will lift all boats 2m6s.

The Future of AI Valuation and Market Growth

• The valuation of companies like Anthropic, which is based on its ability to access compute, and the role of competitors like XAI, which is starting to sell compute, will be important factors in determining the value of these companies and the AI industry as a whole 2m6s.
• The current state of AI adoption is still in its early days, despite trillions of dollars of value already being created, and it feels like the first, second, or third innings of a game 10s.

Challenges in AI Adoption and ROI Measurement

• AI is already creating real business value, but companies are struggling to determine their return on investment (ROI) for their AI spending, with some going over budget and trying to wrestle with the costs 2m6s.
• The concept of "token maxing" has been a major topic of discussion, where companies are trying to maximize their AI usage, but this has led to some crazy headlines, such as Uber reportedly blowing through their annual budget in a few months 4m30s.
• Some companies, like Meta, have taken measures to limit AI usage, such as shutting off licenses or removing leaderboards, and there is a quote that "when a measure becomes a target, it no longer works as a measure" 6m20s.
• The focus has shifted from "token maxing" to determining ROI, which is the right measure, especially as companies deploy more AI agents and need to measure accurate spend per agent, use case, team member, or user 8m10s.
• There are different approaches to measuring ROI, including standalone companies that are working on metering and tracking AI usage, and some companies, like Factory, have announced their own solutions, such as a model router, to help with ROI measurement 12m0s.
• Enterprises are still in the process of figuring out their return on investment (ROI) for artificial intelligence (AI) and optimizing costs in real-time, with companies like Ramp providing cost management and spend intelligence, including tokenomics and token spend 10s.
• The top 1% of firms that are most invested in AI are spending $7,500 per employee per month on AI, which is less than half of what they would pay for a software engineer, with the median spend being around $12 per employee per month 2m6s.
• Most companies are not spending as much on AI as the top 1%, with the top 10% spending only $611 per employee per month, and the majority of companies are only at 1-2% of AI adoption 4m6s.

Strategies for AI Integration and Optimization in Enterprises

• To fully utilize AI, companies need to map out their workflows, identify time-intensive tasks, and create AI systems that can augment their value drivers, which requires creative tinkering and often comes from the founder down 6m6s.
• Nvidia and Merkor have recently stated that they are spending more on AI than on human employees, but according to Ramp's index reports, this is not yet the case, with AI spend still being lower than employee salaries 1m42s.
• Enterprises are likely operating at less than 50% capacity when it comes to AI, leaving significant room for growth, and the top 1% of firms tend to mix and match multiple frontier models and platforms to access cheaper open source models 10s.
• The use of AI models and platforms varies depending on the team type, with some teams using Claude, Perplexity, or Chatbt, while engineering teams may use Codecs, Cloud Code, and other tools, and enterprises often buy multiple tools and are still in the test and see mode 42s.
• Startups like Ramp and Factory that provide services for managing AI ROI, spend, and model routing may have a standalone path as a long-term venture bet, particularly in the observability play where a third-party tracks cost and ROI 2m6s.

AI Workflows, Tools, and Platform Usage

• The concept of an AI native workday or HCM platform that manages both employees and agents is also an interesting angle, and ROI could be a good starting point, but the exact outcome is still to be determined 2m6s.
• The use of AI is becoming ubiquitous, and successful tracking of ROI will be crucial, making it a natural fit for applications that can provide this functionality, and companies like Factory with model routing products have a promising outlook 2m6s.
• Many companies will likely develop ROI products that can be integrated alongside their core products, providing a comprehensive view from different angles 10s.
• To facilitate adoption and improve tool usage, companies like OpenAI and Anthropic are sending forward deployed engineers to work directly with enterprise organizations, helping them automate specific workflows 42s.

Forward Deployed Engineers and AI Adoption

• Forward deployed engineers, also referred to as a "Trojan horse," can help customers successfully adopt and use AI products while also identifying gaps and areas where AI can be further applied, allowing for quick time-to-value and potential development of new products 2m6s.
• The role of forward deployed engineers is similar to that of consultants, but with a focus on AI adoption and workflow automation, and they can provide valuable insights to product teams, enabling rapid development cycles 2m6s.
• The introduction of forward deployed engineers is not causing anxiety among engineers, but rather excitement, as they see it as an opportunity to complement and enhance their workflows, potentially leading to 10x or 100x productivity gains 42s.

AI Automation, Job Displacement, and Workflow Enhancement

• Enterprises are still figuring out how to drive value with AI and determine their return on investment, with the goal of automating tasks and saving human time for more valuable interactions 10s.
• The idea of using AI to automate workflows and potentially replace human jobs is being explored, with some companies working on creating AI-driven standard operating procedures that can learn from human behavior and capture ingrained knowledge 2m6s.
• These AI systems can sit on a computer and watch what humans are doing, effectively acting as a digital full-time equivalent employee, and then provide insights on how to improve processes 2m6s.
• Despite the potential benefits of AI, there are also concerns around security and the potential for AI systems to disrupt traditional workflows, with some preferring the current full-time equivalent model 2m6s.

Connecting with Tiffany Luck

• Tiffany Luck can be connected with online through LinkedIn, where she responds to most of her messages, and also on other platforms such as X 4m42s.