This guide is written for people aiming at revenue, GTM and AI automation work, where portfolios matter more than in most fields because the roles are new and credentials are thin. The logic behind it is covered in proof of work vs. credentials.
Why a practice portfolio works
Employers still look for experience. In the World Economic Forum's Future of Jobs Report 2025, 81% of businesses said they expect to keep using work experience to assess skills, more than any other method. Skills assessments came second at 48%.
You cannot manufacture a job history. But what employers are really trying to learn from "experience" is whether you can do the work in a realistic context. A practice project that mirrors a real business problem answers a big part of that question.
There is also a structural reason. LinkedIn's Economic Graph Research Institute estimates that looking at skills rather than prior job titles expands the pool of candidates for AI roles by 8.2x globally. But research from the Burning Glass Institute and Harvard Business School found that many companies that dropped degree requirements did not meaningfully change who they hired. A wider door only helps you if a reviewer can quickly see your skills. A portfolio is how you make them visible.
The rule that makes a portfolio credible: problems, not tools
Most beginner portfolios are tool tours: "Here is a Clay table. Here is an n8n workflow. Here is a ChatGPT prompt." A reviewer learns that you opened the tools. They do not learn whether you can create value.
Flip it. Every project starts with a business problem and ends with a number. The tool is just the middle. This is the core of MitHub's Revenue Reverse Engineering method: follow the money backwards from the payment to find where revenue leaks, then build.
The MitHub Practice Project Canvas
Before you build anything, fill in these seven lines. If you cannot, the project is not ready.
- Business type. For example, "a hypothetical dental clinic network with 8 locations."
- Revenue leak. Where money is lost: slow follow-up, bad data, no visibility.
- Who feels it. The role that suffers: front desk, sales rep, owner.
- Baseline. How it works now, in numbers, even if simulated.
- Smallest system. The minimum you can build to change the baseline.
- Measure. What number moves and how you will check it.
- Evidence. What a reviewer will be able to open.
7 practice project ideas in revenue and automation work
These are practice projects. Use simulated data, public information or your own processes, and say so in the write-up. Each one maps to one of the four families of GTM systems from prove value fast: lists and enrichment, agentic AI systems, automated workflows, and data and reporting.
1. Speed-to-lead responder (automated workflows)
- Problem: Inbound leads wait hours for a first reply.
- Build: A form submission triggers a workflow that validates the contact, sends an immediate personalized reply and notifies the right person.
- Measure: Time from submission to first reply, before (manual, simulated) and after.
- Evidence: Workflow export, a recorded test run, a timing log.
2. Lead routing by location (automated workflows)
- Problem: A multi-location business sends every lead to one shared inbox.
- Build: Routing rules by zip code or city to the nearest location, with a fallback when no match exists.
- Measure: Share of test leads routed correctly, including edge cases.
- Evidence: Rules table, 30 test leads, a list of edge cases you found.
3. Ideal customer list with enrichment (lists and enrichment)
- Problem: A sales team prospects from a random list.
- Build: Define an ideal customer profile, pull a list from public sources, enrich it, and score each account.
- Measure: Share of records with usable contact data; how many passed your qualification rules.
- Evidence: The ICP definition, the scored table, the scoring logic written in plain language.
4. CRM data quality audit (data and reporting)
- Problem: Duplicates and missing fields make the pipeline unreliable.
- Build: Take a simulated export of a few hundred records with deliberate errors. Detect duplicates, missing phones and invalid emails, and propose ownership rules.
- Measure: Error rate before and after cleaning.
- Evidence: Before and after files, the cleaning steps, the ownership rules.
5. Source-to-revenue report (data and reporting)
- Problem: The owner does not know which channel produces paying customers.
- Build: A simple report tracing each paid deal back to its first contact source.
- Measure: Revenue and conversion by source, in simulated data.
- Evidence: Dashboard or sheet, the data model, one decision the report would change.
6. AI call or message summarizer (agentic AI systems)
- Problem: Reps do not log what happened in conversations.
- Build: Take sample transcripts you write yourself, generate structured summaries (outcome, objection, next step) and push them to a CRM-style table.
- Measure: Accuracy against a manual reading of the same transcripts.
- Evidence: Prompt, input samples, output table, a list of errors the AI made.
7. Objection playbook from conversations (agentic AI systems + diagnosis)
- Problem: Every rep handles objections differently.
- Build: Classify objections from sample conversations and draft a playbook with a response for each.
- Measure: Share of sample objections the playbook covers.
- Evidence: Classification table, the playbook, notes on what you would validate with a real team.
Pick three that match the roles you want. A candidate aiming at GTM engineering might choose 1, 3 and 5. Someone aiming at AI automation might choose 1, 6 and 7.
How to document each project
Use the four-part structure from your case study:
- Situation. The business, the leak and who feels it. Two or three sentences.
- Path. What you decided, what you built and what you rejected. Show judgment, not just steps.
- Result. Baseline versus after, with the limits of your test stated plainly.
- Evidence. Links a stranger can open in under five minutes.
Add one more line most people skip: what I would do with a real client. It shows you understand the difference between a practice environment and production.
Where your portfolio sits on the proof ladder
Practice projects are not the top of the ladder, and you should not pretend they are. In MitHub's proof ladder, a documented practice project is an applied artifact. Above it are real results, verified results and repeated results.
The next step up is to get a real person involved:
- Offer to rebuild a small process for a local business, a nonprofit or a friend's company, with their permission.
- Ask a practitioner to review one project and tell you where it would break.
- Use one of your systems in your own work and track it for a month.
Each of these moves one project from "I built it" to "someone used it." That is a meaningful jump in credibility.
Common mistakes
- Using a real company's private data. Never do this without permission. Simulated data is fine and expected.
- Inflating results. A 30-lead test does not prove a 40% increase in revenue. Say what you tested.
- Too many projects. Ten shallow projects signal dabbling. Three deep ones signal capability.
- Hiding that it is practice. Honest labeling is a sign of judgment. Getting caught is the end of the conversation.
- Tool-first write-ups. Lead with the problem and the result, not the software list.
A portfolio checklist
Before sharing, check each project:
- Starts with a business problem, not a tool
- Clearly labeled as a practice project if it is one
- Has a baseline and a measured result
- Has evidence a stranger can open in five minutes
- States what you did personally: executed, directed, designed or owned
- Contains no confidential or personal data
- Explains what would change in a real deployment
Where to go next
If you do not yet know which problems to practice on, the free foundations of the Revenue Reverse Engineering faculty walk through six chapters, each ending in one proof. If you want to see how the portfolio fits into a longer path, read learn, build, prove, earn. Experience is something you accumulate. Evidence is something you can start producing this week.