Points to ponder…
- Perplexity Token Issues: User frustration growing due to pricing; only three Pro searches daily for $20/month.
- Competition Pressures: Early adopters migrating to alternatives like OpenAI and Claude due to limitations and costs.
- Industry Pricing Shift: Rapid token usage changes affecting costs; companies exploring flat-rate pricing to stabilise expenses.
- AI Training Resources: Google offers modular courses on AI; ongoing learning essential for staying current in legal tech.
- Community Engagement Caution: Proposals to open feedback sessions; anonymity crucial for maintaining confidentiality in legal discussions.
Notes from webinar fireside chat
Perplexity Platform Evaluation and Market Position
Perplexity faces growing pains with token consumption limits impacting user experience and market positioning. (07:09)
- Token usage issues cause user frustration as Malcolm Pearson experienced token depletion without receiving images, revealing a loop where tokens were consumed but no output delivered.
- The free tier allows unlimited quick searches but only three Pro searches daily; Pro is priced at $20/month with unlimited pro searches and advanced LLM access.
- The Max tier at $200/month offers agentic AI capabilities and priority support but is expensive for typical users.
- Perplexity focuses on enterprise plans to boost profitability amid increasing token usage costs.
- Perplexity aggregates data from multiple LLMs but has faced criticism for its data scrubbing and aggregation methods, indicating potential compliance or quality concerns.
- The platform launched “Perplexity Computer” as an agentic AI tool for workflows, requiring the Max plan.
- “Perplexity Comet,” an AI-focused web browser, aims to enhance browsing but remains untested by users.
- User migration trends show early adopters moving to alternatives like OpenAI and Claude for faster, better responses, reflecting competitive pressure.
- Feedback from Reddit and community forums suggests initial enthusiasm has waned due to pricing.
- Delegate 1 valued Perplexity’s broad LLM access at $20/month, making it cost-effective compared to separate subscriptions.
- User experience varies by plan, with Pro sufficient for legal use cases according to delegate 2, who has not encountered token limits on the Pro plan but notes the high cost of Enterprise.
- They use Perplexity’s history and workspace features to track legal matters, enhancing workflow continuity.
Token Pricing Challenges and Industry Implications
The rapid and unpredictable token consumption is causing cost concerns and forcing re-evaluation of AI pricing models. (19:17)
- Microsoft capped Claude code usage due to runaway AI spend, signalling big tech’s control over AI cost risks.
- Microsoft’s move aims to promote internal AI tools like Copilot, reducing reliance on third-party LLMs.
- Uber exhausted its entire 2026 AI coding budget in just four months, showing real-world token cost spikes.
- This accelerated usage, three times faster than expected, highlights unsustainable token burn rates in active developer teams.
- It would seem token pricing models need to shift from pay-as-you-go to flat-rate plans to stabilise costs.
- There is speculation that future pricing will offer unlimited tokens for fixed monthly fees to avoid user churn.
- Delegate 1 notes tokens now burn for document analysis alone, increasing costs compared to earlier models.
- Some companies consider building in-house AI models as annual token costs for multiple users exceed alternatives.
- This shift could impact AI vendors’ revenue streams and market dynamics.
- AI coding costs may surpass human developer salaries, raising ROI questions on AI adoption.
- Malcolm stresses that AI coding must pay for itself to justify token expenses.
AI Training and Learning Resources
Learning tools and courses are emerging to help users and lawyers keep pace with AI developments. (25:27)
- Google offers modular AI and LLM courses at skills.google.com, providing quick, structured learning paths.
- Malcolm recommends these courses for legal professionals wanting foundational AI knowledge. More below…
- Claude provides broad AI training resources, as highlighted by Delegate 2, useful for deeper understanding.
- These official vendor resources may complement external courses and help users better leverage AI platforms. More below…
- “AI for Dummies” style guides are impractical due to rapid AI evolution, making online and modular courses more viable.
- Malcolm notes that printed materials quickly become outdated given the pace of AI advancements. More below…
- Ongoing learning is vital to keep up with AI tools’ capabilities and limitations, ensuring users make informed decisions.
Legal Industry AI Adoption and Use Cases
Legal professionals are experimenting with AI but face operational and educational challenges. (24:16)
- Delegate 3’s AI rollout is delayed due to internal system issues, indicating technical or organizational barriers.
- This delay impacts the timing of AI integration in workflows.
- Delegate 4 uses AI tools like Gemini but seeks simple, practical guidelines to improve usability.
- His request for beginner-friendly AI education underlines a knowledge gap in legal tech adoption.
- Delegate 2 benefits from using Perplexity Pro for legal matters, leveraging persistent workspaces and history to track case-related AI interactions.
- This use case suggests AI can enhance legal research and case management if implemented carefully.
- Participants agree that AI’s usefulness depends on balancing cost, token limits, and platform features for legal contexts.
- The high cost of enterprise plans may limit widespread adoption in smaller firms.
Community Engagement and Confidentiality Considerations
Opening AI feedback sessions publicly could increase engagement but requires careful privacy management. (31:53)
- Malcolm proposes anonymising feedback content to open sessions to a wider audience, aiming for greater interaction and knowledge sharing.
- This could attract more participants and broaden the community while preserving confidentiality.
- Delegate 4 supports trialing open sessions to gauge response, suggesting a low-risk approach.
- Delegate 1 emphasizes strict confidentiality to protect law firm identities, noting that shared knowledge differs from firm representation.
- Maintaining trust is critical given the sensitivity of legal discussions.
- Delegate 3 and Delegate 5 agree openness could enrich information flow, supporting transparency without compromising privacy.
- The group consensus favours cautious expansion of access with safeguards, balancing inclusiveness and professional discretion.
Action items
Malcolm Pearson
- Review previously published AI learning resources and provide links alongside skills.google.com AI courses to assist Frank with beginner-friendly materials (27:16)
- Verify and anonymize sensitive details in AI for Lawyers session feedback before trialing public release of anonymized summaries to attract engagement while preserving confidentiality (31:53)
- Publish anonymised AI for Lawyers session feedback summaries within one to two days after the meeting to share insights and encourage wider participation (28:46)
Malcolm’s Action points…
We spoke about A Dummies Guide to AI, and low and behold, this is what came through my Google Alerts…
https://www.snowflake.com/wp-content/uploads/2024/01/Generative-AI-and-LLMs-for-Dummies.pdf
The Google AI learning is at:
https://www.skills.google/
Anthropic – Claude Courses
https://anthropic.skilljar.com/
