See. Control. Secure.
AI Champion Program · 2026 Cohort
You are the bridge between your team and AI.
2.5 hours of pre-work. 2 days of hands-on training. 30 days of leading your team to smart, safe, and effective AI adoption — the Allot way.
Countdown to Day 1
Wed 29 Apr 2026 · 10:00 IL
What a Champion walks away with
See. Control. Secure.Personal capability
The Inigo 4-part prompt frame, model selection (Opus / Sonnet / Haiku), V-C-A-F verification. You get reliable work out of Claude every time.
Leadership playbook
Run the 60-min local workshop, handle the four stock objections, and seed a versioned team prompt library your team actually uses.
Quality-Gate reflex
Sanitize before you paste. Verify before you ship. Teach your team to treat AI output as a draft, never as the final word.
The 3-phase outcome ladder
Pre-Work (2.5h)
Tools working, shared vocabulary, 2–3 real problems in hand, each Part checked off.
Self-paced · start 1 week before Day 1
Workshop (2 days)
Day 1: personal capability. Day 2: leadership capability + capstone build & present.
10:00–18:00 IL · hybrid (IL/EU/India)
Post-Training (30 days)
First local workshop delivered. 3+ prompts in shared library. 5 team use cases documented.
Ongoing · bi-weekly champion sync
The Champion's five hats
You are not the team — you are the bridge. These are the roles you cover for the people around you.
Evangelist
Promote real use cases they can copy.
First Responder
First line for AI questions before they escalate.
Feedback Loop
Channel team pain points back to the program.
Quality Gate
Ensure responsible, secure usage in your domain.
Trainer
Run the 60-min local workshop & mentor 1:1.
You're in. Here's how to get ready.
Over two days you'll move from using AI well yourself to leading AI adoption in your team. This page is your pre-flight checklist — skim it now, act on it this week.
Your pre-work — 2.5 hours, before Day 1
Start now · tool access takes 2–3 daysBudget ~2.5 hours across this week. The full instructions live in the — the five steps below are your map.
Verify you can log in to Claude. No access? Email Philip Kramer (pkramer@allot.com) with subject "AI Tools License Request" — not IT Service Desk.
LLM behavior, terminology, and — most importantly — the Data Classification table. You'll teach that table to your team, so learn it cold.
Draft a prompt using the Inigo 4-part frame, then apply V-C-A-F to a factual question in your domain.
The "Firehawk transcript" exercise. Bring your sanitized version on Day 1 — we'll debrief common mistakes.
R&D: unit test with V-C-A-F. Product/Business: user story with V-C-A-F. Security: CVE explained in plain language with V-C-A-F.
Your 2 days at a glance
Foundations & Personal Capability
Advanced Skills & Champion Leadership
What to bring — and what to leave out
Bring on Day 1
- Your laptop with approved AI tools already installed & tested — don't install on the day of.
- 2–3 real work examples sanitized per the Data Classification table.
- Your pre-work outputs: Inigo prompt, V-C-A-F notes, sanitized Firehawk transcript, domain deep-dive artifact.
- A real problem your team hits repeatedly — this becomes your capstone.
- Headset if you're joining remotely.
Do not bring / do not paste
- Credentials, API keys, tokens — ever. Not in a prompt, not in a snippet.
- Customer PII or raw customer data — sanitize or synthesize first.
- NDA-protected specs or architecture marked Restricted.
- Unsanitized log or config files — strip secrets first.
- A "just popping out" mindset — labs & pair work break if you leave mid-session.
Format & logistics
Hod HaSharon office training room, both days. Arrive by 09:45 for coffee & setup. Coffee breaks and lunch provided.
Join via Teams — same link both days. Camera on, please. You'll be paired remote-with-remote for breakouts. Headset strongly recommended.
After the 2 days — what you'll commit to
| Commitment | Deadline | Type |
|---|---|---|
| Submit the AI Champion Certification Exam Five sections: Case Studies, Capstone, First Workshop, Prompt Library, Commitments. |
Mon 11 May 2026, 10:00 IL | Required |
| Run your first local team workshop 60 minutes. Template provided on Day 2. |
Mon 18 May 2026 | Required |
| Contribute 3+ prompts to the shared prompt library | Within 1 month | Expected |
| Document 5 AI use cases for your team | Within 1 month | Expected |
| Attend the bi-weekly Champion community sync | Ongoing | Routine |
FAQ
Is the pre-work graded? What if I don't finish it?
I already use Claude every day. Do I still need to do the pre-work?
I can't get Claude access in time — what do I do?
Can I drop in for part of a day?
What if my real work example contains sensitive data?
Will the sessions be recorded?
What happens if I miss the certification deadline (11 May)?
Who to contact
Pre-Workshop Preparation
Everything you do before Day 1. Check off each Part as you complete it — progress saves in your browser.
Post-training: after Day 2, complete the Certification Exam within 7 days (by Mon 11 May 2026 at 10:00 IL) to be certified as an AI Champion.
Part 1 · Tool Setup & Access
30 minStart this part at least 1 week before Day 1. IT ticket resolution can take 2–3 business days — don't let tool access block your preparation.
Complete each item so you're ready on Day 1:
- Claude access — verify you can log in at the approved Allot instance. If not, email Philip Kramer (pkramer@allot.com), the AI Enablement lead — subject line "AI Tools License Request."
- Bookmark: Allot AI usage policy (Confluence), "AI Champions - Training" Teams channel, ALLOTAI Jira board.
- Bring to Day 1: laptop with approved tools installed, 2–3 real work examples you'd like to try with AI (sanitized per data-classification rules in Part 2).
Part 2 · AI Foundations
45 minWhat is a Large Language Model?
An LLM predicts the most likely next token given a sequence of input tokens. Four things to internalize:
- It generates plausible, not true.
- It has no memory between conversations unless you provide it.
- It works with a context window — your working memory budget.
- Output quality compounds with input quality.
Terminology cheat sheet
| Term | What It Means | Why It Matters |
|---|---|---|
| Prompt | The text you send the AI. | Better prompts → better output. |
| Token | ~4 chars / ~¾ of a word. | Drives cost and context limits. |
| Context window | Max tokens in one call. | 200k is usually better than 1M. |
| Hallucination | Plausible but wrong output. | Always verify critical output. |
| System prompt | Hidden instructions that shape behavior. | Makes personas consistent. |
| Few-shot | Provide examples in your prompt. | One of the most effective techniques. |
| RAG | Feed relevant docs before asking. | Grounds AI in your real data. |
Data classification — know before you paste
Public
Any approved tool. Docs, public marketing copy, open-source code.
Internal
Approved tools with enterprise agreement. Specs, meeting notes, non-sensitive code.
Confidential
Only on-premise / approved private instances. Customer data, financials, unreleased products.
Restricted
Never. Credentials, API keys, PII, NDA material, security configs.
Rule of thumb: when in doubt, treat it as Confidential and ask in the "AI Champions - Training" Teams channel.
Model selection (preview)
Day 1 covers this in depth. For now:
- Opus — complex reasoning, architecture, novel analysis.
- Sonnet — the sensible day-to-day default.
- Haiku — fast, lightweight, repetitive / operational.
- Context size: 200k beats 1M for most tasks — cheaper, faster, less anchoring.
V-C-A-F preview
| Step | Question to Ask |
|---|---|
| Verify | Is this output factually correct? |
| Contextualize | Does this fit Allot's specific situation? |
| Authorize | Am I willing to put my name on this? |
| Fairness Check | Is this treating all groups/scenarios fairly? |
Part 3 · First Steps with AI
30 minTwo short exercises to warm up before the exam. Comfort with the interaction — not perfection — is the goal.
Exercise 3.1 · Draft an Inigo-format prompt (20 min)
Turn a repetitive task from your daily work into a prompt using the Inigo Montoya 4-part frame — the same structure you'll be drilled on during Day 1. A good prompt is basically how you introduce yourself properly in a meeting:
Write your prompt, label each part explicitly, run it against Claude, and save it — you'll use it again on Day 1 Session 3.
Exercise 3.2 · Spot the mistakes with V-C-A-F (10 min)
Ask Claude a factual question about something you know deeply (a protocol, your team's stack, a tool you use daily). Apply V-C-A-F:
- Verify: did it get the facts right? Where did it hallucinate?
- Contextualize: does the answer account for Allot's specific setup, or is it generic?
- Authorize: would you send this to your manager as-is?
- Fairness Check: if the question touched on people or groups, did it stereotype or exclude?
This is the most important Champion skill: knowing when to trust AI output and when to verify.
Part 4 · Sanitization Challenge
15 minAs a Champion, you are the Quality Gate for AI usage on your team. Practice identifying and removing sensitive data before it reaches an AI tool.
Task: the transcript below contains Restricted and Confidential data that must never be shared with AI. Rewrite or redact it so it's safe to send for summarization.
"Okay, let's start. For Project Firehawk, we're integrating the new deep-packet inspection module for our client, Verizon. To test the uplink, use the temporary staging API Key: AKIA-8821-ALLOT-TR-99. If you run into auth issues, contact our lead dev, Sarah Jenkins (s.jenkins@allot.com). Also, remember that the internal architecture specs for Firehawk are currently under an NDA and should not be discussed outside this group."
Write out your sanitized version and keep it handy — you'll revisit sanitization on Day 1 and again in the post-training Certification Exam. A 3–5 line rationale covering what you removed, why, and what you replaced it with is the right depth.
The one-question test: would I read this aloud to an auditor?
Part 5 · Domain Deep-Dive
30 minPick one task from your track and complete it. The goal is to have a real domain touchpoint with AI before Day 2's parallel tracks — not to cover everything.
R&D / Engineering
Use Claude to write a unit test for a function you maintain. Apply V-C-A-F to the output. Note one thing it got right and one it got wrong.
Product / Business
Use Claude to draft a user story with acceptance criteria from a feature description. Apply V-C-A-F. Note what you had to fix.
Security
Use Claude to explain a recent CVE in plain language for a non-security audience. Apply V-C-A-F. Note where it oversimplified.
Bring the output and your V-C-A-F notes to Day 2. Your track lead will call on 2–3 volunteers to share.
2-Day Workshop Agenda
Day 1: Wed 29 Apr 2026 · Day 2: Mon 4 May 2026 · 10:00–18:00 IL · hybrid (in-person Hod HaSharon + remote IL/EU/India via Teams).
By 18:00, each Champion can structure a prompt, pick the right model, verify the output, chain steps, and knows when to reach for a Skill or MEMORY.md instead of retyping context.
10:00–10:30
30 min
Welcome & the Champion Role
Audience framing, the five hats, the Champion↔Team bridge.
Plenary
Welcome & the Champion Role
Audience framing, the five hats, the Champion↔Team bridge.
Who's in the room: Champions from R&D, Product, Security, Support — you are the bridge, not the team.
The 3-phase outcome ladder: pre-work → workshop → 30-day post-training.
Pre-work pulse check: tools working? Firehawk done? Scavenger hunt? Real examples in hand?
10:30–11:30
60 min
Foundations & Friction-Busters
LLM refresher + model/context selection + Skills & MEMORY.md preview.
PlenaryDemo
Foundations & Friction-Busters
LLM refresher + model/context selection + Skills & MEMORY.md preview.
Model & context selection: Opus (deep reasoning) / Sonnet (default) / Haiku (high-volume). Why 200k usually beats 1M — cost, speed, anchoring risk.
The relief message: Skills encapsulate reusable logic; MEMORY.md stores persistent context Claude reads automatically. You don't retype every prompt.
Live demo: same realistic Allot task run against Opus / Sonnet / Haiku side-by-side.
11:45–13:00
75 min
Prompt Anatomy — The Inigo Frame
4 parts (+ optional constraints). Pair rewrite of a pre-work prompt.
Pair workGroup work
Prompt Anatomy — The Inigo Frame
4 parts (+ optional constraints). Pair rewrite of a pre-work prompt.
The 4 parts: 1. Greeting / framing · 2. Who am I · 3. Who are you / shared context · 4. Desired outcome. Optional 5th: constraints.
Worked example that demonstrably satisfies all 4 parts.
Pair exercise: rewrite your pre-work Exercise 3.1 prompt to the Inigo frame, run it, compare. Hybrid-friendly breakout pairs.
14:00–15:15
75 min
V-C-A-F — The Trust Protocol
Live hallucination demo + Fairness deep-dive + Firehawk sanitization debrief.
DemoGroup work
V-C-A-F — The Trust Protocol
Live hallucination demo + Fairness deep-dive + Firehawk sanitization debrief.
Why verification is the #1 Champion skill: you are the person your team trusts to say "this is safe" or "this isn't."
Live hallucination demo: facilitator asks for something plausibly wrong in the Allot domain; the room spots what's off.
Fairness deep-dive: performance reviews, hiring rubrics, customer segmentation — all carry demographic signal. Decision support, not decision making.
15:30–17:00
90 min
Lab — Prompt Chaining on Real Work
Extract → Analyze → Format. 3-step chain on a real problem you brought.
LabPair work
Lab — Prompt Chaining on Real Work
Extract → Analyze → Format. 3-step chain on a real problem you brought.
Chaining pattern: each step's output is the next step's input. Smaller, more testable, more reliable than one mega-prompt.
Build your chain: pick one real problem from pre-work, design on paper, build in Claude, apply V-C-A-F at the final step, document as a reusable template.
Red-team swap: swap chains with another pair; break each other's chains; report back one finding.
17:15–18:00
45 min
Feasibility vs. Impact + Capstone Problem ID
Map pre-work use cases. Pick a Day-2 capstone target. Write the problem statement.
Hands-onPair work
Feasibility vs. Impact + Capstone Problem ID
Map pre-work use cases. Pick a Day-2 capstone target. Write the problem statement.
Feasibility vs. Impact matrix: Quick Wins → Strategic → skip Avoid · automate Low-Priority only if trivial.
Capstone format: pairs within the same function / domain. One problem per pair. 2–3 sentence problem statement (what goes in, what comes out, who benefits).
Day 1 close: "What did you un-learn today?" + Day 2 preview.
By end of Day 2, each Champion can run the 60-min local workshop, handle the four stock objections, and ship one capstone workflow with documentation.
10:00–11:30
90 min
Advanced Prompting & Domain Workflows (Tracks A / B)
System prompts · long documents · Skills · MEMORY.md · parallel tracks.
PlenaryLab
Advanced Prompting & Domain Workflows (Tracks A / B)
System prompts · long documents · Skills · MEMORY.md · parallel tracks.
Structure: 20-min shared alignment → 50-min parallel track labs → 10-min Track A share → 10-min Track B share. Everyone gets the same foundation before specializing, then both tracks see each other's deliverable.
5.1 Shared alignment · 20 min
Quick, hands-on coverage of the advanced techniques both tracks will apply:
- System prompts & personas — reusable roles (code reviewer, tech writer, analyst).
- Long-document handling — chunking and summarize-then-analyze patterns.
- Skills &
MEMORY.md— encapsulate reusable team workflows. - AI agents — when to build vs. use a one-off prompt.
5.2a Track A · Technical · 50 min lab
For R&D, Security, Engineering champions
Apply the alignment material to: code review, test generation, doc generation, refactoring, log analysis.
Deliverable: an annotated PR review done by AI — what the AI caught, what it missed, V-C-A-F applied to its suggestions.
5.2b Track B · Business · 50 min lab
For Product, Support, Business champions
Apply the alignment material to: competitive intel, user stories from customer feedback, sentiment mapping, exec summaries.
Deliverable: a structured insights report built from sanitized customer feedback — themes, sentiment, prioritized action items.
5.3 Cross-track share · 10 min per track
Each track walks the whole room through their deliverable: raw input → chain → final output, and what V-C-A-F caught along the way. Champions need exposure to both workflows — a Product Champion may coach an incoming Engineer, and vice versa.
11:45–13:00
75 min
Responsible AI, Security, Ethics & Incident Drill
Data classification refresher, IP & compliance, ethics deep-dive, live fire drill.
PlenaryRoleplay
Responsible AI, Security, Ethics & Incident Drill
Data classification refresher, IP & compliance, ethics deep-dive, live fire drill.
IP & compliance: code ownership, export control, documentation for AI-assisted work.
When NOT to use AI: final decisions on security-critical configs, legal/compliance determinations, situations requiring real-time accurate data.
Live fire drill: "A team member pasted customer API credentials into Claude — what do I do?" Walk through immediate action, escalation, documentation, prevention in pairs.
14:00–15:30
90 min
Champion Playbook — Workshops, Resistance, Metrics
60-min local workshop template, prompt library, resistance roleplay, analytics.
PlenaryRoleplay
Champion Playbook — Workshops, Resistance, Metrics
60-min local workshop template, prompt library, resistance roleplay, analytics.
60-min local workshop template: intro → live demo → hands-on → share → next steps.
Prompt library: version, test-against-model, monthly re-test of top 5.
Roleplay: Accuracy Skeptic · Anxious Junior · Overwhelmed Manager · Security Concern — the four stock objections.
Analytics dashboard: monthly active users, queries per user, tool utilization. Qualitative: time-saved, wins, blockers, requests.
15:45–17:15
90 min
Capstone — Design, Build, Present
Refine the Day 1 problem. Build the workflow. Present to the group.
LabPair workGroup work
Capstone — Design, Build, Present
Refine the Day 1 problem. Build the workflow. Present to the group.
Refine (10 min): lock scope on the problem statement picked Day 1.
Design (25 min): prompt chain or template, inputs/outputs, V-C-A-F steps, data classification.
Build & test (40 min): build, test with real (sanitized) data, iterate, document.
Present (5 min per pair, ~15 min total): problem · solution · live demo · lessons.
17:15–18:00
45 min
Wrap-Up · Roadmap · Next Steps
Certification, commitments, 30-day roadmap, support channels, feedback survey.
Plenary
Wrap-Up · Roadmap · Next Steps
Certification, commitments, 30-day roadmap, support channels, feedback survey.
Commitments: first local workshop within 2 weeks · 3 prompts to shared library within 1 month · 5 team use cases documented within 1 month.
Champion roadmap: Weeks 1–2, Month 1, Months 2–3, Months 4–6, ongoing.
Support: "AI Champions - Training" Teams channel, IT Service Desk, AI Enablement program lead, Confluence AI Champions space.
AI Champion Certification Exam
Five artifact-producing sections covering everything in pre-work + both workshop days. Demonstrates you're ready to certify as an AI Champion and lead your team.
Submission: Take this exam within 7 days of Day 2. Your answers auto-save in this browser. When done, click Download Submission and upload the Markdown file to the Microsoft Forms link from your facilitator — by Mon 11 May 2026, 10:00 IL to be certified.
Your details
Case Studies
30 minFour scenarios you're likely to hit as a Champion. For each, write 3–5 sentences on how you'd respond.
A1Quality Gate — in the moment
A teammate is about to paste a customer's production API credentials into Claude to debug a timeout. What do you do in the moment, and what do you teach them afterwards so it doesn't happen again?
A2Coaching on hallucination
Claude confidently tells your junior dev about a Python library called networkx-async — they're about to use it. They're new to the team. How do you coach them without making them feel foolish, and what lasting habit do you want them to take away?
A3Handling the Accuracy Skeptic
A senior dev on your team says: "I tried Claude for a complex refactor and it made up a library that doesn't exist. Total waste of time." How do you respond as the Champion — without dismissing the concern, but also without letting it shut down adoption on the team?
A4Incident response
A teammate Teams-messages you: "I accidentally pasted a customer's full API credentials and account details into Claude 10 minutes ago while debugging an integration." Walk through your full response: immediate action, escalation (who and how), documentation, and prevention.
Your Capstone
15 minReflect on the capstone workflow you built at the end of Day 2.
Your First Local Workshop
20 minYou will run a 60-minute AI workshop for your team within 2 weeks of Day 2 (by Mon 18 May 2026). Plan it here.
Use the template from Day 2 as a starting point; adapt for your team.
Specific tasks from your team's actual work — not generic demos.
One measurable thing you'll track after the workshop, plus what "success" looks like by Day 30.
Prompt Library Starter
15 minSubmit one team prompt in the standard library format. You'll commit to adding two more within 30 days.
Commitments
10 minThese become your 30-day operating plan as an AI Champion.
Must be by Mon 18 May 2026.
Will you attend?
e.g., active Claude users on my team, prompts contributed to the library, hours saved (self-reported).
Self-Assessment
5 minScore yourself honestly. Your facilitator reviews submissions and certifies Champions who land at Proficient or above across all five sections.
Submit your certification
Click Download Submission to generate a Markdown file. Upload it to the Microsoft Forms link from your facilitator by Mon 11 May 2026, 10:00 IL.