Artificial intelligence & machine learning investment firm

An investment firm puts money into companies; an AI focus means judging data, models and teams.

• The main stages are seed, early and growth, and each has a different level of risk and information.
• Money comes from limited partners, and managers owe them honest, regular reports.
• AI companies differ: tools, applications, data, services. Judge whether the data or the model is the real asset.
• Write a one-page thesis: what we back, why now, what we avoid.
• Read one real company's public filings or website and list five claims you can check and five you cannot.
• Ask what a customer would do if the company vanished tomorrow. A product people could not easily replace is a stronger sign than a clever model.
• Define the sectors you know in Canada, such as health, logistics, energy or agriculture, and say what you would need to learn before you judge a company in each.

Activity. Write a thesis and test it against three real companies' public information.

Standard to meet. Give reasons and exclusions for the thesis.

Raising and investing money is regulated.

• Provincial securities regulators govern who may raise funds, give advice and sell investments; exemptions exist for some investors.
• Managers may need registration; do not act before checking.
• Anti-money-laundering rules can require identifying investors and sources of funds; check which rules apply to you.
• Marketing must be balanced; never promise returns.
• Check that you can lawfully see the information you use; do not trade on private information.
• Keep every message to investors accurate and dated; if a figure is an estimate, label it as an estimate.
• Rules differ by province and by type of investor. Check the provincial rule and ask a licensed adviser before any pitch.

Activity. Build a rules table for a small fund and ask a licensed adviser to review.

Standard to meet. Check that no row is guessed and opinion is kept.

Due diligence means checking claims.

• Check the data: source, rights to use, quality and bias.
• Check the model: accuracy on held-out data, cost to run, drift over time.
• Check the team and customers: paid pilots count more than letters of intent.
• Write findings in a short memo with a rating and open questions.
• Ask what the company would do with money, by when, and how you would know it worked.
• Compare accuracy on a fair test: examples the model has never seen, drawn the way real customers' data is drawn.
• Ask for cost to serve one customer per month and the price they pay, and compute the gross margin.

Activity. Review one public AI company and write a two-page memo.

Standard to meet. Support each claim with evidence or mark it unverified.

Understand how money flows.

• A fund of 10,000,000 with a 2% yearly fee pays its managers 200,000 a year.
• Returns come from a few big winners while most investments return little, so model that spread.
• Model three cases for the fund: poor, expected and strong.
• Be clear about fees and carried interest, and show investors how each is calculated.
• Compare fund costs over ten years with a fee table, not a feeling.
• Multiple of money = money back divided by money invested. A result of 1.0 means investors only got their money back.
• Note that a multiple before fees is higher than what investors keep; show both before and after fees.

Activity. Build a three-case return table.

Standard to meet. Show fees and assumptions.

After investing, help and monitor.

• Hold a quarterly review of each company's revenue, cash, hires and risks.
• Reserve money for follow-on rounds in the companies that are working.
• Set clear triggers for extra support or for exit.
• Report to investors in a fixed format every quarter.
• Record why you made each decision on the day, so you can learn later.
• Give each company a status: on plan, behind or needs help, with a reason and a date to look again.
• Count months of cash left (cash divided by monthly spending) and flag any company under 9 months.

Activity. Build a quarterly dashboard for a fictional portfolio.

Standard to meet. Give every company a status and next step.

Good deals come from trust.

• Network with universities, accelerators and founders.
• Give founders fast, clear answers.
• Offer help beyond money: introductions and advice.
• Track deal flow through five stages: seen, met, diligence, offer and closed.
• Keep a list of founders you met and notes for follow-up, with their consent.
• Say no kindly and quickly, with one honest reason. Founders remember how you treated them.
• Ask each founder how they found you and track which source brings the best deals.

Activity. Run a deal funnel for 4 weeks (public companies or case studies).

Standard to meet. Record the funnel.

Invest responsibly.

• Screen each company for harmful uses and biased data.
• Declare every conflict of interest in writing.
• Respect federal and provincial private-sector privacy rules in portfolio companies; check the rule that applies in the province.
• Publish an investment policy that investors and founders can read.
• Write the policy for what you do if a founder behaves badly.
• Ask each company how it tests for unfair results across groups of people, and ask to see the test.
• Keep a conflicts register: who owns or advises what, and who left the room when a decision was made.

Activity. Write a one-page responsible investing policy.

Standard to meet. Include screens and a review step in the policy.

Practise with a paper portfolio.

• Pick five public companies and write a thesis for the group.
• Track them for 60 days with dated notes.
• Compare the result with a simple benchmark.
• Write a review of what the thesis got right and wrong.
• Keep your paper fund simple: no more than five names.
• Write the entry reason and the price for each name on day one, before the results arrive.
• State clearly that a paper fund is practice; it is not advice and nothing is bought.

Activity. Run the paper fund and present the review.

Standard to meet. Review separates luck from skill.

L1: Digital aware learner. You can define: name the main terms, people and rules of artificial intelligence & machine learning investment firm. Focus module: Understand how investment in AI works (Understand, 6 h). An investment firm puts money into companies; an AI focus means judging data, models and teams. Key points: The main stages are seed, early and growth, and each has a different level of risk and information. Money comes from limited partners, and managers owe them honest, regular reports. AI companies differ: tools, applications, data, services. Judge whether the data or the model is the real asset. Write a one-page thesis: what we back, why now, what we avoid. Read one real company's public filings or website and list five claims you can check and five you cannot. Ask what a customer would do if the company vanished tomorrow. A product people could not easily replace is a stronger sign than a clever model. Define the sectors you know in Canada, such as health, logistics, energy or agriculture, and say what you would need to learn before you judge a company in each.

Speaker notes. Teach: open with the worked example. Thesis: 'We back applied AI for Canadian clinics and logistics at seed stage; we avoid pure model-building without data access.' A test on 3 companies might find 2 that fit and 1 that fails the data rule, and the reason for the failure is written beside it. Then set the activity: Write a thesis and test it against three real companies' public information. Mark against the standard: Give reasons and exclusions for the thesis. Check at this level: What are the stages of company investing? Model answer: Seed, early and growth. Level note: Start here if new. Evidence: a one-page glossary and a visit or interview note.

L2: Support assistant. You can describe: explain how artificial intelligence & machine learning investment firm works from first contact to result. Focus module: Know the securities rules (Apply, 6 h). Raising and investing money is regulated. Key points: Provincial securities regulators govern who may raise funds, give advice and sell investments; exemptions exist for some investors. Managers may need registration; do not act before checking. Anti-money-laundering rules can require identifying investors and sources of funds; check which rules apply to you. Marketing must be balanced; never promise returns. Check that you can lawfully see the information you use; do not trade on private information. Keep every message to investors accurate and dated; if a figure is an estimate, label it as an estimate. Rules differ by province and by type of investor. Check the provincial rule and ask a licensed adviser before any pitch.

Speaker notes. Teach: open with the worked example. Row: 'Raise a fund from friends | provincial securities regulator | check exemption and registration | before any pitch | lawyer's written opinion'. A fund table with 8 rows, each with an owner and a date, shows at a glance which of the 8 are still open before a first pitch. Then set the activity: Build a rules table for a small fund and ask a licensed adviser to review. Mark against the standard: Check that no row is guessed and opinion is kept. Check at this level: Who regulates raising funds? Model answer: Provincial securities regulators. Level note: Evidence: a flow chart and a two-minute spoken explanation.

L3: Junior developer or analyst. You can apply: carry out a routine task in artificial intelligence & machine learning investment firm to a written standard. Focus module: Evaluate an AI company (Apply, 10 h). Due diligence means checking claims. Key points: Check the data: source, rights to use, quality and bias. Check the model: accuracy on held-out data, cost to run, drift over time. Check the team and customers: paid pilots count more than letters of intent. Write findings in a short memo with a rating and open questions. Ask what the company would do with money, by when, and how you would know it worked. Compare accuracy on a fair test: examples the model has never seen, drawn the way real customers' data is drawn. Ask for cost to serve one customer per month and the price they pay, and compute the gross margin.

Speaker notes. Teach: open with the worked example. Claim '95% accuracy' tested on the company's own test set; independent sample shows 84%, a gap of 11 points. Memo rating drops from strong to watch. If a customer pays 1,500 a month and serving them costs 400, gross margin is 1,100 (about 73%). Then set the activity: Review one public AI company and write a two-page memo. Mark against the standard: Support each claim with evidence or mark it unverified. Check at this level: What is due diligence? Model answer: Checking a company's claims and risks before investing. Level note: Evidence: a supervised task record that meets the standard once.

L4: Developer, consultant or builder. You can analyse: diagnose why an artificial intelligence & machine learning investment firm result missed its target and fix it. Focus module: Model returns and fees (Evaluate, 8 h). Understand how money flows. Key points: A fund of 10,000,000 with a 2% yearly fee pays its managers 200,000 a year. Returns come from a few big winners while most investments return little, so model that spread. Model three cases for the fund: poor, expected and strong. Be clear about fees and carried interest, and show investors how each is calculated. Compare fund costs over ten years with a fee table, not a feeling. Multiple of money = money back divided by money invested. A result of 1.0 means investors only got their money back. Note that a multiple before fees is higher than what investors keep; show both before and after fees.

Speaker notes. Teach: open with the worked example. Invest 500,000 across 10 companies at 50,000 each. Outcomes: 6 return nothing, 3 return 50,000 each, 1 returns 400,000 (8x). Total back = 150,000 + 400,000 = 550,000, which is 1.1x before fees. If the 400,000 winner is missed, the total is 150,000, only 0.3x. Fees on a 10,000,000 fund at 2% for 10 years are 2,000,000. Teaching figures in CAD. Then set the activity: Build a three-case return table. Mark against the standard: Show fees and assumptions. Check at this level: Why do fees matter? Model answer: They reduce investor returns every year. Level note: Evidence: a fault or gap analysis with a fix and a repeat test. This course targets this level.

L5: Senior specialist or team lead. You can evaluate: judge options for artificial intelligence & machine learning investment firm against cost, risk and quality. Focus module: Source and win deals (Create, 6 h). Good deals come from trust. Key points: Network with universities, accelerators and founders. Give founders fast, clear answers. Offer help beyond money: introductions and advice. Track deal flow through five stages: seen, met, diligence, offer and closed. Keep a list of founders you met and notes for follow-up, with their consent. Say no kindly and quickly, with one honest reason. Founders remember how you treated them. Ask each founder how they found you and track which source brings the best deals.

Speaker notes. Teach: open with the worked example. Funnel: 120 seen, 30 met, 8 diligence, 3 offers, 2 closed. Rates: met 30/120 = 25%, diligence 8/30 = 27%, offers 3/8 = 38%, closed 2/3 = 67%. Overall 2/120 = 1.7%, so about 60 deals seen per closing. Then set the activity: Run a deal funnel for 4 weeks (public companies or case studies). Mark against the standard: Record the funnel. Check at this level (this question is covered in module 2: Know the securities rules): Why never promise returns? Model answer: Rules forbid it and results are uncertain. Level note: Evidence: an options paper with numbers and a recommendation defended to a panel.

L6: Delivery or product manager. You can create: design and pilot a new offer or process in artificial intelligence & machine learning investment firm. Focus module: Handle AI ethics and conflicts (Evaluate, 6 h). Invest responsibly. Key points: Screen each company for harmful uses and biased data. Declare every conflict of interest in writing. Respect federal and provincial private-sector privacy rules in portfolio companies; check the rule that applies in the province. Publish an investment policy that investors and founders can read. Write the policy for what you do if a founder behaves badly. Ask each company how it tests for unfair results across groups of people, and ask to see the test. Keep a conflicts register: who owns or advises what, and who left the room when a decision was made.

Speaker notes. Teach: open with the worked example. Policy line: 'We do not back face-recognition used without consent.' A conflicts register with 5 entries, each with a date and an action, lets an investor see that every conflict was handled. Then set the activity: Write a one-page responsible investing policy. Mark against the standard: Include screens and a review step in the policy. Check at this level: Which kind of data does federal private-sector privacy law cover? Model answer: Personal information held by private organisations. Level note: Evidence: a pilot with measured results over at least 30 days.

L7: Digital director or founder. You can lead: run a team or business in artificial intelligence & machine learning investment firm and develop others. Focus module: Pilot a mock fund (Create, 8 h). Practise with a paper portfolio. Key points: Pick five public companies and write a thesis for the group. Track them for 60 days with dated notes. Compare the result with a simple benchmark. Write a review of what the thesis got right and wrong. Keep your paper fund simple: no more than five names. Write the entry reason and the price for each name on day one, before the results arrive. State clearly that a paper fund is practice; it is not advice and nothing is bought.

Speaker notes. Teach: open with the worked example. Paper portfolio +3% vs benchmark +1%, a gap of 2 points over 60 days; reason noted: luck vs thesis. If one of five names rose 12% and the other four were flat, 12 / 5 = 2.4 points came from a single name, which suggests luck. Then set the activity: Run the paper fund and present the review. Mark against the standard: Review separates luck from skill. Check at this level: What separates luck from skill? Model answer: A written thesis tested against results. Level note: Evidence: a record of results over a year and of people you have coached to a higher level.