On February 26, 2026, the Government of Alberta tabled its provincial budget — hundreds of pages of fiscal documents covering a $9.4 billion deficit, $74.6 billion in revenue, $83.9 billion in expenses, and $28.3 billion in capital commitments over three years. Thousands of line items affecting every sector of the Alberta economy, every municipality, every public institution, and every household.
Thirty minutes later, I published albertabudget.ca — a comprehensive analysis website with 25 sector-by-sector breakdowns, 26 audience-specific impact guides, and 40 stakeholder briefing notes. Full analytical pages built from the source documents, cross-referenced against prior-year budgets, with confidence levels on every claim and citations back to the original PDFs.
The tweet was seen over 75,000 times. The implications matter more than the reception — for the government relations industry, for public policy, and for who gets to participate in the conversation that follows a budget.
How budget day works in government relations#
When a provincial or federal government tables a budget, it triggers a scramble. Government relations firms — lobbyists, in plain language — need to analyze the fiscal documents and translate them into actionable intelligence for their clients. A firm representing a major oil producer needs to know what changed in royalty frameworks. A firm representing a hospital association needs to find health capital funding. A firm representing a construction association needs to pull infrastructure line items buried across six ministry business plans.
This analysis is expensive. GR firms in Calgary and Edmonton have teams of analysts who spend budget day reading documents, pulling numbers, and writing client briefings. A five-to-ten page memo explaining what the budget means for a specific client is what clients pay retainers for. A single stakeholder briefing can represent thousands of dollars in billable analysis.
The information asymmetry is the product. GR firms have the expertise and the bandwidth to read the documents quickly. Their clients do not. That asymmetry has existed for as long as governments have published budgets. On February 26, I dissolved it in thirty minutes.
Ninety-two pages of budget analysis, published simultaneously#
albertabudget.ca is a publishing platform that produces the same kind of analysis that GR firms sell — but for every sector, every audience, and every stakeholder type simultaneously.
25 sector analyses covering oil and gas, electricity, data centres, agriculture, forestry, healthcare, education, housing, transportation, and sixteen more — each with budget allocations, prior-year comparisons, risks and opportunities, and cited source documents.
26 audience-specific guides explaining what the budget means for families, farmers, startup founders, nurses, teachers, Calgary residents, seniors, newcomers to Alberta, and eighteen other groups — each translating the fiscal data into the impacts that matter to that audience.
40 stakeholder briefing notes structured the way a GR firm would write them — key numbers, risk assessment, recommended advocacy positions, and suggested talking points for major oil producers, oilfield services companies, data centre developers, hospital boards, teachers’ associations, chambers of commerce, and more.
Every page carries a confidence level and cites its sources. The site discloses that AI was used in the analysis and directs readers to verify against the official government documents.
Multi-agent orchestration for policy analysis#
The system behind albertabudget.ca is a nine-stage pipeline with over fifty specialized AI subagents running concurrently. I built it on the multi-agent orchestration architecture I described in an earlier post — but applied to a real problem with real stakes and a hard deadline.
The pipeline moves through extraction, normalization, parallel analysis, content generation, quality assurance, enrichment, and publishing. The design decisions that make it work are less about the AI and more about the engineering: how you decompose the analysis problem, how you manage shared context so that fifty agents do not contradict each other, how you verify consistency across ninety-two output documents, and how you sequence the work so that the most important content publishes first while the rest of the pipeline is still running.
The hardest part is normalization — building the shared factual layer that every downstream agent references. A fiscal summary agent and a key numbers agent produce the canonical figures for the budget: total revenue, total expenditure, deficit, and the line items that matter for each sector and audience. Every agent that runs after that point draws from the same source of truth. When the healthcare sector page and the audience page for nurses both cite health spending, they cite the same number, because they derived it from the same normalized context. When they do not, a QA stage catches the discrepancy before anything publishes.
The analysis stage is where the parallelism pays off. Twenty-five sector agents and twenty-six audience agents run concurrently, each with a prompt tailored to its domain and the shared fiscal context appended. The output is structured analysis — data that the generation stage turns into readable pages with specific numbers, year-over-year comparisons, and cited sources.
Building this required two decades of domain knowledge in public affairs and political communications, and years of experience building software systems at scale. The AI is the multiplier. The expertise is the thing being multiplied.
What this means for government relations#
A GR firm might have three analysts who spend four hours each producing five client briefings on budget day. That is sixty person-hours for five deliverables. My pipeline produces ninety-two in thirty minutes.
A senior GR professional brings relationships, political context, and strategic judgment that shapes what a client does next — and that work is not going anywhere. But the analytical foundation underneath it — the reading, the number-pulling, the cross-referencing, the structured summarization — is the bulk of budget-day labor, and it is precisely what multi-agent orchestration handles well.
The industry will need to move up the value chain, from analysis to strategy, from translation to action. The firms that see this will use AI to multiply their capacity. The dinosaur firms that do not will be slower than the firms that do, and will be left asking “what’s the frequency, Kenneth?” while the asteroid comes for them.
AI and the future of policy analysis#
Government relations is a specific case of a broader pattern. Governments routinely produce massive policy documents — budgets, regulatory frameworks, legislative packages, committee reports, environmental assessments — that are effectively inaccessible to most citizens. The documents are public, but they are dense, technical, and voluminous. Most people — including most journalists, most business owners, and most elected officials outside the relevant committee — do not have the time or the background to read a provincial budget and understand what the numbers mean.
The result is an information asymmetry in democratic governance. The people most affected by a budget decision learn about the impacts days or weeks later, filtered through media coverage that may not address their specific situation. albertabudget.ca compressed that timeline to thirty minutes — plain-language guides, sector analyses, and stakeholder briefs published simultaneously from the same source documents the lobbyists are reading.
For twenty years, I have been building technology that shapes public affairs — online networks that changed how political conversations formed in Canada, data infrastructure that made election analysis accessible beyond party war rooms, media intelligence systems that compressed response times from days to minutes. Dozens of projects, many of which made national news. The through-line is the same argument: technology is not a channel for public affairs — it is the infrastructure. The organizations that treat it as core infrastructure operate at a different speed than those that treat it as an afterthought. albertabudget.ca is that argument applied to policy analysis.
Every organization that touches government policy will need this#
Alberta’s budget is one case. The approach applies to any large structured policy release — federal budgets, regulatory proposals, environmental assessments, omnibus legislation, trade agreements.
This is where the industry is going. Every GR firm, every industry association, every advocacy organization, every newsroom that covers policy will eventually need the ability to ingest a major release and produce audience-specific analysis faster than a team of analysts can read the table of contents. The organizations that build this capability first will define the framing. The organizations that wait will be responding to someone else’s.
The question is not whether to adopt this, but how. You do not get from a PDF to ninety-two verified, cross-referenced analytical pages by prompting a chatbot. The architecture, the prompt engineering, the normalization layer, the QA pipeline, the domain expertise baked into every stage — these take significant experience to build well. The organizations that start now will have a decisive advantage over those that start when their competitors have already shipped.
From budget day to continuous policy intelligence#
albertabudget.ca ran once, on budget day. The next iteration monitors government releases continuously — regulatory gazettes, order-in-council publications, committee proceedings, ministerial announcements — and produces analysis as documents are published.
The broader consequence is about who participates in policy conversations at all. When comprehensive budget analysis is freely available within an hour of tabling, the people who were never GR clients — community organizations, parent councils, local business associations — suddenly have access to the same analytical foundation as the people paying five-figure retainers. The conversation broadens. The accountability is focused.
I have spent twenty years building technology that changes who gets to participate in public affairs. I have built the system that does this for a provincial budget. I am now working with organizations that want to build their own. If that is you, get in touch.
