Semilattice logo with icon depecting overlapping sets.Semilattice logo with icon depecting overlapping sets.

User simulation

for AI workflows

Add simulated user and market data into your AI workflows to test content, personalise UX, and tune decision logic.
Help me think through GTM for our new API feature
U
AI
Sure! Is there anything in particular you're focused on?
Pricing... what kind of pricing model do developers generally prefer?
U
AI
I'll simulate a developer audience to explore their pricing preferences.
AI
Using Semilattice
Simulating
userSimulation.ts
const result = await semilattice.answers.simulate({
  population_id: "Developers,
  answers: {
    question: "Which pricing model you prefer?",
    question_options: { question_type: "single-choice" },
    answer_options: [
      "Usage-based",
      "Monthly subscription",
      "Monthly with annual discount"
    ]
  }
});
Use Cases

Test content, empower agents, and personalise UX

The Semilattice API lets you simulate user opinions and behaviour programmatically, injecting new data into your AI workflows and business logic. Use it to improve generated content, give agents new information about the user, and tailor UI flows for specific user profiles.
// Generate and test copy until we find a winner
async function findWinningCopy(semilattice, populationId) {
  while (true) {
    
  // Generate 3 copy variations
    const copyLines = generateCopyAlts();
    
    // Test with developers
    let result = await semilattice.answers.simulate({
      population_id: populationId,
      answers: {
        question: "Which message resonates most with you?",
        question_options: { question_type: "single-choice" },
        answer_options: copyLines
      }
    });
    
    // Poll until prediction is complete
    while (result.data[0].status !== "Predicted") {
      await new Promise(resolve => setTimeout(resolve, 1000));
      result = await semilattice.answers.get(result.data[0].id);
    }
    
    // Return winner if >50%, otherwise try again
    const percentages = result.data[0].simulated_answer_percentages;
    const winner = Object.keys(percentages).find(answer => percentages[answer] > 0.5);
    if (winner) return winner.answer;
  }
}
Copy TestingImprove the effectiveness of generated content by testing and iterating lines of copy with simulated audiences.
Workflow
  1. Generate 3 lines of copy for a Developer audience
  2. Call the Semilattice `answers` endpoint to test which lines the Developers audience find most compelling
  3. If no line gets >50%, repeat with new lines of copy
Integration Contexts
Marketing AutomationWriting ToolsDesign Tools
// Generate and test copy until we find a winner
async function findWinningCopy(semilattice, populationId) {
  while (true) {
    
  // Generate 3 copy variations
    const copyLines = generateCopyAlts();
    
    // Test with developers
    let result = await semilattice.answers.simulate({
      population_id: populationId,
      answers: {
        question: "Which message resonates most with you?",
        question_options: { question_type: "single-choice" },
        answer_options: copyLines
      }
    });
    
    // Poll until prediction is complete
    while (result.data[0].status !== "Predicted") {
      await new Promise(resolve => setTimeout(resolve, 1000));
      result = await semilattice.answers.get(result.data[0].id);
    }
    
    // Return winner if >50%, otherwise try again
    const percentages = result.data[0].simulated_answer_percentages;
    const winner = Object.keys(percentages).find(answer => percentages[answer] > 0.5);
    if (winner) return winner.answer;
  }
}
Population Models

Use pre-trained population models or create your own

Every Semilattice population model is tested for accuracy with transparent metrics returned from the API to inform decision confidence. Get started with one of our ready-made models or create custom models for specific user segments with a small sample of data.
US Gen Z & Millennials
87.36%
A model of US adults aged 18-44, based on the 2023 National Public Opinion Reference Survey.
Designers
85.45%
A model of designers based on the 2023 uxtools.
Front-End Developers
88.50%
A model of Front-End Developers based on the 2024 Stack Overflow Developer Survey.
Data Engineers
85.19%
A model of Data Engineers based on the 2024 Stack Overflow Developer Survey.
Full-Stack Developers
90.99%
A model of Full-Stack Developers based on the 2024 Stack Overflow Developer Survey.
Back-End Developers
85.54%
A model of Back-End Developers based on the 2024 Stack Overflow Developer Survey.
DevOps Specialists
88.69%
A model of DevOps Specialists based on the 2024 Stack Overflow Developer Survey.
Data Science Community
82.74%
A model of the Data Science community—data scientists, ML engineers, data analysts, research scientists, statisticians, ...
Enterprise Application Developers
89.75%
A model of Desktop & Enterprise Application Developers based on the 2024 Stack Overflow Developer Survey.
Data Scientists
87.72%
A model of Data Scientists & Machine Learning Specialists based on the 2024 Stack Overflow Developer Survey.
Embedded Application Engineers
82.50%
A model of Embedded Application & Device Engineers based on the 2024 Stack Overflow Developer Survey.
Engineering Managers
88.07%
A model of Engineering Managers based on the 2024 Stack Overflow Developer Survey.
US Adults
88.26%
A nationally representative model of US adults based on the 2023 National Public Opinion Reference Survey.
Developers - AI focus
87.63%
A model of Developers with a focus on AI tooling based on the 2024 Stack Overflow Developer Survey.
Developers
92.71%
A model of Developers based on the 2024 Stack Overflow Developer Survey.
Mobile Developers
85.42%
A model of Mobile Developers based on the 2024 Stack Overflow Developer Survey.
Pricing

Try for free

Start for free and scale as you need.

Free

API v1
Explore our API for free
$0/month
  • 50 free predictions
  • Ready-to-use Population Models
  • API access
Get API Key

Play

API v1
Build tools with no commitment
$1per prediction
Everything in Free plus:
  • Custom Population Models
  • Premium support
  • Pay as you go
Get API Key

Launch

API v1
Launch tools at scale
$399/month
Everything in Play plus:
  • 500 predictions per month (rolling over)
  • 25% discount on additional predictions
Talk To Us
Benchmarking Results

87% accurate on unknown questions

All population models created on Semilatatice are tested for accuracy and our simulation engines are benchmarked regularly. Our latest benchmarking on hundreds of questions from 12 randomly selected target audiences showed an average accuracy of 87.21%.
Benchmarked Population Models
Fantasy Football PlayersBenchmarked
Test Accuracy86.76%
UK Adults with Inflation ConcernsBenchmarked
Test Accuracy92.01%
NHS Healthcare WorkersBenchmarked
Test Accuracy89.16%
DevOps SpecialistsBenchmarked
Test Accuracy88.69%
Data Scientists & ML SpecialistsBenchmarked
Test Accuracy87.72%
DesignersBenchmarked
Test Accuracy85.45%
American Adults (Panel Wave 90)Benchmarked
Test Accuracy88.34%
US AdultsBenchmarked
Test Accuracy88.26%
Developers - AI focusBenchmarked
Test Accuracy87.63%
Students in Part-Time Degree ProgramsBenchmarked
Test Accuracy85.99%
American Adults (Panel Wave 91)Benchmarked
Test Accuracy84.51%
DevelopersBenchmarked
Test Accuracy92.71%
US Gen Z & MillennialsBenchmarked
Test Accuracy87.36%
Engineering ManagersBenchmarked
Test Accuracy88.07%
Embedded Application & Device EngineersBenchmarked
Test Accuracy82.50%
Back-End DevelopersBenchmarked
Test Accuracy85.54%
Data EngineersBenchmarked
Test Accuracy85.19%
Consensually Non-Monogamous IndividualsBenchmarked
Test Accuracy86.30%
Nigerian CitizensBenchmarked
Test Accuracy87.19%
UK Voters in 2019 ElectionBenchmarked
Test Accuracy83.86%
Residents of Somerville, MABenchmarked
Test Accuracy90.52%
Data Science CommunityBenchmarked
Test Accuracy82.74%
Mobile DevelopersBenchmarked
Test Accuracy85.42%
Desktop & Enterprise Application DevelopersBenchmarked
Test Accuracy89.75%
Front-End DevelopersBenchmarked
Test Accuracy88.50%
Full-Stack DevelopersBenchmarked
Test Accuracy90.99%