Slide 1: AI Agents, Human Insights & Productivity
For personal use only AI Agents, Human Insights & Productivity
Slide 2: Disruptive technology must improve productivity
“If any technology is be truly disruptive it has to For personal use only show it dramatically improves productivity, simplifies processes, reduces costs or improves the environment” This presentation will give a live of example of an AI Agent disrupting legacy processes 2
Slide 3: When genAI became widespread we had a perception problem
When genAI become widespread we had a perception problem! For personal use only As a company specialising in translation and localisation, people wondered how our business model would survive if AI can do translation, despite the fact we had been using AI in translation for more than a decade. “Ge
Slide 4: Where can we add value?
Where can we add value? For personal use only We went on a journey to understand how our existing AI and human processes, would add maximum value combined with the latest LLM technology What were the things that genAI didn’t do well? How do we utilise our asset base with the new opportunities genAI
Slide 5: Exploring the landscape
Exploring the landscape For personal use only So we built a genAI fact-checker and went on tour We exhibited at AI global conferences Paris, London, New York to gauge interest in verification of AI What we found out: Interest in genAI for Quality Evaluation in workflows more than content generatio
Slide 6: Technology doesn't change the core function being performed
Technology doesn’t change the core function being performed For personal use only Technology makes something more efficient but is often still achieving the same thing DVDs to Streaming – still big screen entertainment at home Letters to Email – still person to person communication Electric cars – s
Slide 7: AI enabled a new solution for the industry, so we built one
AI enabled a new solution for the industry, so we built one For personal use only With the AI fact-checking technology in place, we started to look at how we could use it as an AI Verification engine We developed a quality estimation engine using our own internal corpus of language data and language
Slide 8: Automated Post Editing – LLMs talking to each other
Automated Post Editing – 3 LLMS talking to each other For personal use only 8
Slide 9: How pre-AI business models worked
How pre-AI business models worked For personal use only 1 2 Machine MT 10,000 words x Translation Human 10,000 words 0.0005c word = $5 review Review at 10 hours x 1,000 $50/hour = $500 words/hour = 5c per word Total cost: $505 Cost per Hourly or per word of MT word rate *rates and costs are basic ex
Slide 10: VERIFY AI Agent Process Flow
VERIFY AI Agent Process Flow Yes Use it For personal use only Is LLM Cycle Source content (APE) Content good enough 3 Can an No LLM make it good enough LLM generates Trained LLM content for Quality 1 Evaluation No Yes 2 Send for human Use it verification 4
Slide 11: Verify quality evaluation
Verify quality evaluation For personal use only 11
Slide 12: VERIFY AI Agent Charging Model
VERIFY AI Agent Charging Model For personal use only 1 2 3 4 LLM generates Send for human Trained LLM content for Quality LLM APE verification Cycle Evaluation Hourly rate Token rate Token rate 2 Automated @ $50/hour 1 based based specific postediting x 1 Hour = general LLM LLM Token rate 3 $50 1,00
Slide 13: Orchestrate AI workflow builder
Orchestrate AI workflow builder For personal use only Customers can configure rules and workflows for the AI translation process based on their budgets and required quality outcomes across language pairs and subject domains. 13
Slide 14: Key components to our Verify AI Agent
6 key components to our Verify AI Agent For personal use only 1. Content generation through LLM 2. Quality Evaluation & Judgement through SLM 3. Refinement of output through multiple LLMs 4. Action to human verification 5. Configuration Engine 6. Changing the payment model to token based charging 14
Slide 15: AI Agent for Gap Analysis
AI Agent for Gap Analysis For personal use only The same process and tools used for AI Verification can be applied to other solutions such as audit gap analysis 2 4 1 3 LLM compares Trained LLM for Quality Multi-LLM Human documents/content Evaluation specific to Automated insight the use-case Review
Slide 16: AI Gap Analysis
16 For personal use only AI Gap Analysis
Slide 17: Summary of AI agents
Summary of AI agents For personal use only 1. Decision on being a model maker, taker or shaper 2. LLMs combined with smaller, more specific models 3. Every disruption comes with a change in pricing 4. Pricing moves to tokens/characters 5. Quality Evaluation replaces human decision making 6. Actions
Slide 18: Thank you
For personal use only Thank you grant@straker.ai www.straker.ai
Slide 19: Disclosure Statement
Disclosure Statement For personal use only This presentation is given on behalf of Straker Limited ASX:STG (Company number NZ: 1008867 / AU: ARBN 628 707 399) Information in this presentation: ● Is for general information purposes only, and is not an offer or invitation for purchase, or recommendati