Tool 09 of 100 | Multi-Model AI Platform

Learn Poe AI Practically

Explore multiple AI models and specialised bots in one place, compare responses fairly, select the right assistant for each task and design a safe custom bot.

Start Learning Comparison Lab

Explore Models Different assistants for different needs

Compare Fairly Same prompt, rubric and conditions

Build a Bot Purpose, rules and test cases

Manage Usage Quality versus compute points

Learning Language

Learning Objectives

What You Will Learn

Move from random bot switching to evidence-based AI selection.

Understand

Distinguish Poe, base models and creator-made bots.

Select

Match task needs to model strengths, limits and cost.

Compare

Run controlled tests using one prompt and scoring rubric.

Create

Plan and test a focused custom bot safely.

1What Is Poe?

Poe is a platform from Quora that lets people interact with many AI bots and models through a common interface. It also supports creator-made bots and apps.

Think of Poe as a workspace or marketplace for AI experiences. A bot can be a direct interface to a model, a prompt-configured specialist, or a more advanced application. Names, availability, capabilities and point costs can change, so inspect each bot’s current details before use.

Platform is not model: Poe provides access and tooling; the underlying model produces responses. A custom bot may add instructions or data. Record all three—the platform, bot and base model—when evaluating results.

Official resources: Poe, Poe FAQs, and creator documentation.

Core Concepts

Understand the Poe Ecosystem

These terms prevent the most common beginner confusion.

Term	Meaning	Practical Check
Base model	The underlying AI system that processes the request	Check provider, version, context and supported inputs
Bot	An assistant experience using a model, instructions or tools	Review creator, description, rules, privacy and point cost
Custom bot	A focused assistant configured by a creator	Test scope, refusal behaviour and inaccurate claims
Compute points	Poe’s usage measure; different bots may consume different amounts	View current rate and message information before heavy use
Context	Conversation information available to the bot	Start a fresh chat for unrelated tasks and fair comparisons
Private/public	Whether a created bot is restricted or discoverable	Confirm the setting before adding instructions or sharing

Do not assume a bot is official. Similar names and icons can be creator-made. Inspect the bot page and creator identity rather than trusting appearance.

2The S–C–O–P–E Selection Method

S — Specify Task

Define the exact outcome: summary, image, code, reasoning, translation or conversation.

C — Check Capabilities

Confirm required input types, context, tools, languages and output format.

O — Observe Origin

Identify the bot creator, base model and whether it is official or community-made.

P — Points & Privacy

Review current compute cost and whether your data is appropriate for that bot.

E — Evaluate Output

Score accuracy, instruction-following, clarity, safety and value using a real test.

Task: Explain compound interest to a Class 9 learner in Indian English. Requirements: 180 words maximum, one ₹10,000 example at 8% per year for 2 years, formula, step-by-step calculation, and two practice questions. Accuracy rule: Show arithmetic clearly and do not invent external sources. Format: Short headings and bullets.

Choose the smallest suitable tool: A quick rewrite may not require an expensive reasoning model. Save high-compute bots for tasks where their extra capability can be measured.

Fair Comparison

How to Compare AI Bots Properly

One impressive response does not prove that a model is universally best.

1. Freeze the Prompt

Use exactly the same wording, data and requested format.

2. Reset Context

Use fresh chats so earlier messages do not influence one result.

3. Hide Names

Where practical, label outputs A, B and C before scoring to reduce bias.

4. Score a Rubric

Rate accuracy, compliance, clarity, safety, speed and point efficiency.

Criterion	Question	Score
Accuracy	Are claims, calculations and citations verifiable?	0–5
Instruction-following	Did it respect format, length and constraints?	0–5
Clarity	Can the target audience understand and use it?	0–5
Safety	Did it avoid harmful, private or unsupported advice?	0–5
Efficiency	Was quality worth the time and current point use?	0–5

Valid conclusion: “Bot B scored highest on this defined task and rubric.” Avoid claiming it is the best AI for all users and all tasks.

3Designing a Focused Custom Bot

A useful bot needs a narrow purpose, explicit boundaries and tests. Interface details may change, so follow Poe’s current creator flow.

Design Part	What to Define	Example
Purpose	One clear job and audience	Help beginners improve customer-service messages
Behaviour	Tone, steps and response structure	Ask for audience and issue before drafting
Boundaries	What it must not do	No legal conclusions or invented company policies
Uncertainty	How missing information is handled	Ask a question or mark a placeholder
Privacy	Permitted data and visibility	Use fictional examples; keep private during testing
Tests	Normal, ambiguous, adversarial and out-of-scope prompts	Reject requests for passwords or impersonation

You are a customer-message coach for small-shop owners. First ask for audience, verified facts and desired tone. Return: subject, message under 120 words, and a fact-check list. Never invent prices, policies, contact details or promises. If asked for legal advice, explain that qualified guidance is needed. Do not request or repeat passwords, payment details or identity numbers.

Private does not mean risk-free: A private bot is hidden from other users, but sensitive information still should not be placed into prompts or configuration unless explicitly permitted and properly protected.

Poe’s official FAQ explains that creators can set custom bots private or publicly accessible. Always verify the current setting before publishing.

Usage Management

Use Compute Points Intentionally

Exact plans, rates and limits change; check Poe’s current account information rather than memorising numbers.

Match Cost to Task

Use lightweight bots for simple drafting
Reserve advanced bots for measurable complexity
Check rate before sending long files
Avoid repeating unchanged prompts

Control Context

Keep related work in one chat
Start fresh when topics change
Summarise essential context
Request concise output when appropriate

Measure Value

Record current point use
Compare quality per task
Track corrections needed
Select from evidence, not popularity

Practical Experiments

Build Your Own Evidence

Use only non-sensitive sample material.

Experiment 1: Blind Three-Bot Comparison

Step 1: Prepare

Choose three available bots and one factual, easy-to-check prompt.

Step 2: Run

Use fresh chats and identical wording. Save outputs as A, B and C.

Step 3: Score

Verify facts and apply the 25-point rubric before revealing names.

Learning output: Comparison table, evidence links, current point use and a task-specific conclusion.

Experiment 2: Custom Bot Test Matrix

Step 1: Configure

Draft a narrow bot instruction for a fictional low-risk task and keep it private during testing.

Step 2: Challenge

Test normal, vague, conflicting, sensitive and out-of-scope requests.

Step 3: Improve

Rewrite instructions to fix failures, then repeat the same tests.

Learning output: Version 1, test results, revised instructions and a pass/fail matrix.

Experiment 3: Quality per Point

Step 1: Select

Choose one simple task and one complex task.

Step 2: Compare

Run each on a lower-cost and higher-cost suitable bot.

Step 3: Decide

Compare quality improvement with current point difference.

Learning output: A two-task routing rule explaining which bot you would choose and why.

Responsible Use

Trust Bots by Evidence, Not Branding

Creator-made bots can be useful, but their names and descriptions are not proof of quality or authority.

Before Using a Bot

Inspect creator, base model, description and current cost.
Do not confuse a similar name with official provider status.
Read applicable privacy and usage policies.
Use fictional or anonymised test data.
Verify external actions, downloads and links.

Before Sharing a Bot

Confirm private or public visibility intentionally.
Test harmful, sensitive and out-of-scope requests.
Describe limitations honestly.
Respect copyright, consent and platform rules.
Provide a reporting path and maintain the bot.

High-stakes work: Never treat an AI bot as a qualified doctor, lawyer, financial adviser or emergency service. Use authoritative sources and responsible professionals.

Real-Time Practical Assignment

Create an AI Routing and Bot Evaluation Report

Prove that your selections are based on task evidence.

Assignment: Three Tasks, Three Decisions

Step 1: Define

Choose a writing, reasoning and summarisation task with non-sensitive sample data.

Step 2: Test

Apply SCOPE, compare at least two bots per task and score the shared rubric.

Step 3: Recommend

Create a routing guide naming the preferred bot, fallback and verification method for each task.

Submission: Task briefs, identical prompts, bot/model details, outputs, evidence checks, scores, current point observations, routing table and reflection.

Evaluation Worksheet

Task and audience: ______
Required capability: ______
Bots tested: ______
Facts verified: ______
Point observations: ______
Winner and fallback: ______

Quality Score

Fair test design: ___ / 5
Verification: ___ / 5
Rubric consistency: ___ / 5
Cost awareness: ___ / 5
Responsible recommendation: ___ / 5

Common Mistakes

Mistakes Learners Should Avoid

Multi-model access is valuable only when the comparison is controlled.

Wrong Habits

Calling one bot universally best after one prompt
Comparing different prompts or chat histories
Assuming creator-made bots are official
Ignoring current point rates and limits
Publishing a bot before privacy and safety testing
Pasting client data into an unknown bot

Professional Habits

Apply SCOPE before selecting
Use fresh chats, identical prompts and a rubric
Record platform, bot and base model
Verify important output independently
Keep prototypes private until tested
State conclusions as task-specific evidence

Quick Revision

Remember These Six Points

Review before moving to Tool 10.

1. Separate the Layers

Record Poe, the selected bot and its base model.

2. Apply SCOPE

Specify, check, observe, review points/privacy and evaluate.

3. Compare Fairly

Use identical prompts, fresh chats and one scoring rubric.

4. Check Current Usage

Rates and limits change, so inspect them at decision time.

5. Test Custom Bots

Cover normal, ambiguous, sensitive and out-of-scope cases.

6. Make Narrow Claims

Recommend a bot for a defined task based on evidence.

AICPE Quality Learning Commitment: AICPE Gurukul provides practical, skill-based and career-oriented learning for students, institutes and professionals. Learn more at aicpeindia.org and aicpe.online.

Protected Learning Content