AI-native startups like Lio leverage multi-agent systems to tackle complex, end-to-end enterprise workflows, challenging traditional incumbents.
AI-native startups are gaining an advantage by managing the entire arc of a process, not just the final entry, moving beyond the distribution-centric models of traditional software.
Incumbent software companies often struggle to evolve beyond their existing products, which are limited to specific records and fail to capture the 'work behind the scenes'.
Traditional ERP systems only record final outcomes, such as a final price for aluminum, often masking the true complexity of procurement.
The actual procurement process involves extensive behind-the-scenes work, including hundreds of emails, multiple stakeholder meetings, and detailed engineering analyses.
Startups have an opportunity to address this 'invisible' labor performed outside of software records, a gap that incumbent systems currently fail to track.
For example, an ERP might show an 8K price for aluminum, but it won't show the supplier's initial pushback asking for 10K.
True judgment-based AI remains rare among incumbents due to internal incentives and the complexity of integrating such capabilities.
To bridge this capability gap, incumbents are actively partnering with various AI labs, seeking external expertise.
Lio's AI agents span various functions, from basic retrieval and process automation to more complex policy enforcement and principal decision-making roles.
The decision to deploy autonomous agents versus human-in-the-loop systems is determined by factors such as budget, task complexity, and the associated operational risks.
Unlike incumbents who inherently possess customer trust, Lio and similar AI-native startups must actively earn that external trust from their clients.
Lio currently has use cases where agents run fully autonomously, while other situations require human oversight.
Enterprises often lack the capacity to negotiate deals below $50,000, effectively ignoring potential savings in this tier.
Lio automates these 'ignored' negotiations, capturing savings that were previously left on the table due to resource constraints.
Since no negotiation process existed for these smaller deals, the risk associated with using an AI agent for automation is near zero.
Modern procurement tasks extend beyond simple administration, involving complex contract design, financial analysis, and cost engineering.
AI agents feed specialized insights to experts who possess deep knowledge within specific industries.
The procurement process now spans multiple software systems, involving a mix of specialists and generalists to navigate its complexities.
AI agents integrate external data, news, and predictive market indicators to enhance procurement decisions.
Suppliers can be assessed based on reliability metrics, such as rates of missing shipments, providing a more comprehensive view than price alone.
For critical use cases, enterprises prioritize delivery probability over the lowest unit price, ensuring essential goods arrive even if it means a higher cost.
A single AI agent is insufficient for multi-step processes, especially in complex negotiations involving numerous variables.
Multi-agent systems share information in a specific sequence, allowing coordinated completion of intricate tasks.
This system replaces fragmented manual work that typically involves disparate software tools and multiple departments, streamlining end-to-end processes.
Indirect procurement, involving high-volume, low-complexity tasks like MRO parts or marketing services, is suitable for fully autonomous agents.
Direct procurement, particularly for complex builds such as airplanes, involves fewer suppliers but billions in spend, necessitating multi-month expert preparation and human-in-the-loop analysis.
Human engineers analyze real-time indices like aluminum and oil prices to develop informed negotiation strategies for these high-stakes scenarios.
The AI contextualizes price hikes by calculating the actual material cost impact on a specific product, differentiating between the raw material cost increase and the product's overall cost.
This serves as a 'back of house' support system, equipping human negotiators with immediate, data-driven counter-arguments to protect profit margins.
Early-stage businesses that focus on deepening customer relationships and increasing sales naturally build a strong market position.
Owning the end-to-end workflow to create a robust 'data asset' provides the defensibility required for long-term business success, creating a lasting competitive advantage.
A Fortune 500 company attempted to build its own cash collection tool but failed due to poor data quality and a lack of contextual understanding.
Managing multiple ERP systems and cross-functional mappings presents a significant obstacle for internal teams attempting to develop AI tools.
Enterprises are increasingly recognizing that building internal AI tools is not their core competency and often diverts resources from their primary business objectives.
Procurement departments hold more power than sales teams realize in dictating the digital tools that suppliers must use.
If a buyer mandates a supplier to use a specific AI agent, it creates a unified coordination platform that benefits both parties.
Standardization reduces friction in non-price tasks like tracking and documentation, streamlining the overall transaction process.
The concept extends to even adversarial parties, such as law firms, who could benefit from shared agent-based coordination for drafting documents and managing issues.
Forward-deployed engineering currently bridges the gap in understanding complex customer data, acting as a crucial intermediary.
Legacy incumbents often struggle with deployment because their implementation teams are an afterthought rather than an integrated feedback loop.
AI enables continuous learning loops, allowing systems to autonomously handle increasingly complex tasks and even automate the deployment process itself.
Customization, traditionally handled by third-party consultants like Accenture, is increasingly being managed through automated software, simplifying and accelerating deployments.
Current tools primarily focus on improving efficiency, but they fail to fundamentally change the fact that much of the work is still conducted through email, Excel, and Microsoft Teams.
Lio aims to offer a cross-departmental perspective, moving beyond specific feature improvements to address the systemic issues within procurement workflows.
The current tools only make existing processes more efficient without altering how people actually work.
Procurement, often considered a 'boring and niche' category, has not seen a significant revolution in two decades.
It is a highly emotional topic for users who are consistently frustrated with existing systems and manual processes.
Achieving just a 1% saving in procurement can have the same profit and loss impact as increasing revenue by 10%, highlighting its financial leverage.
The immense scale of the procurement sector positions it as a potential trillion-dollar business opportunity for innovation.
If you just manage to get like 1% savings it's like equals like 10% of sales that you have to to get the same outcome in your P&L.
Answers come from the transcript, with the exact spot cited.
Want the next article from a16z?
When a16z publishes, we'll write it up like the one you just read and email it to you.
a16z published 8 in the last 7 days.
