# Ad hoc analysis Source: https://docs.concourse.co/concourse-101/common-use-cases/ad-hoc-analysis When a new question comes in, you should not have to wait days for a custom report. Use Concourse to: * Answer questions like "Which customers increased usage after we launched Feature X?" * Break down churn by industry, deal size, or cohort * Run quick cuts for leadership meetings and Slack threads Start with a natural-language question, let Concourse draft a report, then refine until it is ready to share. # Board and investor reporting Source: https://docs.concourse.co/concourse-101/common-use-cases/board-and-investor-reporting Board decks and investor updates are where numbers meet narrative. Concourse helps you ship both. You can: * Reuse the same workbook each quarter instead of rebuilding from scratch * Auto-refresh core metrics from your systems * Generate draft commentary on key movements and trends * Export tables, charts, and narratives into your existing deck format (PowerPoint, PDF, and more) # Monthly close support Source: https://docs.concourse.co/concourse-101/common-use-cases/monthly-close-support ### Turn your monthly close from a scramble into a repeatable workflow. With Concourse you can: * Compare actuals vs budget and prior periods in seconds * Drill into drivers by segment, product, region, or channel * Generate draft variance commentary that is ready for review Example prompts: * "Walk me through the biggest drivers of variance in net revenue vs budget this month." * "Summarize changes in operating expenses by department and flag anything unusual." # Exports and sharing Source: https://docs.concourse.co/concourse-101/core-concepts/exports-and-sharing Concourse is built so insights do not stay trapped in the tool. You should expect to export and share what you find as part of your regular workflow. With Concourse you can: * **Email report summaries** directly to stakeholders so everyone sees the same numbers and commentary. * **Generate and share slide decks** for board, leadership, and team reviews, exporting into formats like Microsoft PowerPoint and Word. * **Export CSV data** generated by Concourse agents so you can do further analysis in spreadsheets, BI tools, or ad hoc models The goal is not to replace every place you work today. Concourse helps you get to trusted answers quickly, then makes it easy to move those answers into the channels and formats your team already uses. # Integrations Source: https://docs.concourse.co/concourse-101/core-concepts/integrations Concourse can work with both connected systems and one-off files. You can: * Connect core systems like your data warehouse, ERP, payroll, expenses, billing, HRIS, and CRM * Drag and drop CSVs for plan files, ad hoc exports, or one-time lists # Memories Source: https://docs.concourse.co/concourse-101/core-concepts/memories **Memories** capture the way your company talks and thinks about the business. You can store things like: * How you define “Active Customer”, “Net Retention”, or “Qualified Pipeline” * Key strategic priorities * Business-unit, region, and segment definitions Concourse uses memories to: * Apply the same definitions every time * Avoid asking you for basic context on every analysis * Generate commentary that sounds like your finance team, not a generic model As you build analyses, Concourse will often **suggest new memories** based on how you describe metrics, segments, or assumptions. You stay in control: you can accept suggestions that should be remembered going forward, or keep them one-off for that workbook only. # Prompt library Source: https://docs.concourse.co/concourse-101/core-concepts/prompt-library Concourse makes it easy to reuse the prompts that work best for your team. You can: * **Save prompts** you like into your own library so you can rerun them without rebuilding the instructions from scratch. * **Organize a prompt library** around your core workflows and metrics so common questions are always a click away. Concourse also includes **pre-built prompt templates** for common finance workflows, such as: * Month-over-month flux analysis * Expense by vendor analysis You can start from these templates, then adapt and save them so they reflect your definitions, segments, and reporting style. # Report Source: https://docs.concourse.co/concourse-101/core-concepts/report Once you get a workbook or analysis in Concourse to a state you want to reuse and compare against in future periods, you can **turn that workbook into a report**. A report: * Preserves the layout, prompts, and commentary style from the original workbook * Becomes a reusable frame of reference for future periods (for example, month-over-month or quarter-over-quarter comparisons) * Can be scheduled or triggered so you are not rebuilding the same analysis from scratch each time You can run the underlying workbook that powers a report: * **On conditions** – for example, if vendor spend for Microsoft (MSFT) increases by more than 20%, or gross margin drops below 40% * **On a schedule** – for example, on the first of every month or every 30 days to support periodic close reviews, board reporting, or other repeatable deliverables In other words, reports turn your best analyses into proactive, reusable workflows so you always have an "analyst" running in the background. # Workbooks Source: https://docs.concourse.co/concourse-101/core-concepts/workbooks A **workbook** is your home base for a specific workflow or question set. Think: * "Month-End Close – December 2025" * "Q1 Board – Metrics & Commentary" * "Weekly Business Review" Inside a workbook you can: * Group related analyses into one place * Keep numbers, commentary, and context together * Build powerful and detailed visuals to easily interpret your data # How Concourse is different from generic chatbots / OOTB LLMs Source: https://docs.concourse.co/concourse-101/getting-started/how-concourse-is-different-from-generic-chatbots Generic chatbots are not connected to your systems of record. They do not know your chart of accounts, metrics, or business model, and they often hallucinate numbers. Instead of relying on a generic chatbot, Concourse stacks domain expertise on top of the best foundation models. We layer finance and operator‑specific knowledge, schemas, and workflows on top of state‑of‑the‑art LLMs so the AI: * Understands common finance and FP\&A patterns out of the box * Knows how to work with real-world data structures from ERPs, CRMs, and warehouses * Prioritizes correctness and clarity over “creative” answers By constraining the model with finance context and your own definitions, Concourse can be both powerful and predictable. ### Transparent by design Most AI tools answer in a black box. Concourse is built to show its work. When you ask a question, Concourse: * Surfaces the **agent’s reasoning steps** so you can see how it approached the problem * Exposes the **exact SQL queries** it runs against your data sources * Lets you **inspect and rerun** those queries so you can validate numbers and tweak the analysis This transparency gives finance teams confidence that: * Numbers tie back to trusted systems and auditable logic * You can always trace a chart or narrative back to the underlying data and query * Disagreements turn into concrete edits to logic or assumptions, not arguments with a black box # Product overview Source: https://docs.concourse.co/concourse-101/getting-started/product-overview Concourse provides custom AI agents that sit on top of your financial stack and turn raw data into clear answers, commentary, and forecasts. Instead of waiting days for a custom report or living in fragile spreadsheets, your team can ask questions in plain language and get trusted auditable results in minutes. Concourse is built for: * Corporate finance and FP\&A teams running monthly close, forecasts, and board reporting * Operators and GMs who need quick answers to “one more question” from leadership * RevOps and analytics teams that want depth without spinning up new dashboards every time With Concourse you can: * **Analyze** – Chat with your financial data, drill into drivers, and iterate quickly * **Report** – Turn numbers into narratives with AI-generated commentary you can edit and reuse * **Forecast** – Run scenarios and projections on live data so decisions are grounded in reality ### How Concourse is different from traditional BI Traditional BI tools are great for static dashboards and pre-defined metrics. They are slow at: * Answering new questions that do not fit existing dashboards * Explaining the “why” behind a number * Helping you ship board narratives or close packages on a tight deadline Concourse is designed for the workflows where BI falls short: * Freeform questions and follow-ups ("what changed and why?", "show this by cohort", "add a forecast") * Fast narrative generation for close reviews, board decks, and investor updates * Direct ownership by finance and operators, without needing a dedicated analytics engineer for every change ### How Concourse is different from generic chatbots / OOTB LLMs Generic chatbots are not connected to your systems of record. They do not know your chart of accounts, metrics, or business model, and they often hallucinate numbers. Instead of relying on a generic chatbot, Concourse stacks domain expertise on top of the best foundation models. We layer finance and operator‑specific knowledge, schemas, and workflows on top of state‑of‑the‑art LLMs so the AI: * Understands common finance and FP\&A patterns out of the box * Knows how to work with real-world data structures from ERPs, CRMs, and warehouses * Prioritizes correctness and clarity over “creative” answers By constraining the model with finance context and your own definitions, Concourse can be both powerful and predictable. ### Transparent by design Most AI tools answer in a black box. Concourse is built to show its work. When you ask a question, Concourse: * Surfaces the **agent’s reasoning steps** so you can see how it approached the problem * Exposes the **exact SQL queries** it runs against your data sources * Lets you **inspect and rerun** those queries so you can validate numbers and tweak the analysis This transparency gives finance teams confidence that: * Numbers tie back to trusted systems and auditable logic * You can always trace a chart or narrative back to the underlying data and query * Disagreements turn into concrete edits to logic or assumptions, not arguments with a black box # Transparent by design Source: https://docs.concourse.co/concourse-101/getting-started/transparent-by-design Most AI tools answer in a black box. Concourse is built to show its work. When you ask a question, Concourse: * Surfaces the **agent’s reasoning steps** so you can see how it approached the problem * Exposes the **exact SQL queries** it runs against your data sources * Lets you **inspect and rerun** those queries so you can validate numbers and tweak the analysis This transparency gives finance teams confidence that: * Numbers tie back to trusted systems and auditable logic * You can always trace a chart or narrative back to the underlying data and query * Disagreements turn into concrete edits to logic or assumptions, not arguments with a black box # How Concourse is different from traditional BI Source: https://docs.concourse.co/concourse-101/product-overview/how-concourse-is-different-from-traditional-bi Traditional BI tools are great for static dashboards and pre-defined metrics. They are slow at: * Answering new questions that do not fit existing dashboards * Explaining the “why” behind a number * Helping you ship board narratives or close packages on a tight deadline Concourse is designed for the workflows where BI falls short: * Freeform questions and follow-ups ("what changed and why?", "show this by cohort", "add a forecast") * Fast narrative generation for close reviews, board decks, and investor updates * Direct ownership by finance and operators, without needing a dedicated analytics engineer for every change # Anchor the scope: time, entities, and cuts Source: https://docs.concourse.co/concourse-101/prompting-best-practices/anchor-the-scope The more clearly you set the boundaries, the less back-and-forth you need. * **Time frame** – "Last 12 months", "Q4 2025", "Year-to-date vs prior year". * **Entities** – "By product line", "for our Enterprise segment only", "for self-serve customers". * **Cuts and filters** – "Exclude one-time credits", "focus on customers with ARR > \$50k", "only show US + Canada". Example (balanced between Ask and Build): "Why did net revenue change in Q3 vs Q2? Break it down by product line and region. Exclude customers with ARR under \$10k." # Call out key definitions and assumptions Source: https://docs.concourse.co/concourse-101/prompting-best-practices/call-out-key-definitions-and-assumptions Concourse can use your memories, but it helps to be explicit when it matters. * "Use our finance definition of Active Customer, not raw signups." * "Treat expansion as upgrades within the same logo only." * "Use GAAP revenue, not billing-based revenue." If you see something off, correct it once: * "When I say Active Customer, I mean customers with at least one invoice in the last 90 days." Concourse can turn that into a reusable memory so future prompts stay consistent. # Choose the right level of structure Source: https://docs.concourse.co/concourse-101/prompting-best-practices/choose-the-right-level-of-structure More structure is not always better. Use the right level of detail for what you are trying to do. * For **quick exploration and data discovery**, keep prompts fast and conversational: * "Show me ARR by customer." * "Which segments drove the change in gross margin last quarter?" This gives the agent more discretion to decide what views and cuts are most useful. * For **repeatable analysis**, use more structured prompts but still leave room for judgment: * "Break down net revenue change in Q3 vs Q2 by product line and region, and call out the top 5 drivers." * "Draft a monthly close summary focused on revenue, gross margin, and opex vs budget." These are great when you know the shape of the answer you want, but do not need every word specified. * For **templates and documentation-level workflows**, go fully structured: * "Create a monthly revenue report with sections for: (1) summary, (2) chart of ARR by segment, (3) table of top 20 customers by change in ARR, (4) drivers analysis with 3 bullets per driver, (5) risks and follow-ups. Output in markdown that I can paste into our board deck doc." Use this when you want something consistent and shareable that you will run many times. Overall guidance: **start simple, refine or template when needed, and only go fully structured when you want something consistent and shareable.** # Iterate instead of over-specifying Source: https://docs.concourse.co/concourse-101/prompting-best-practices/iterate-instead-of-over-specifying You do not need the perfect prompt up front. Think in passes. 1. Start broad (Ask Mode): "Give me a monthly close summary for November, focusing on revenue and gross margin." 2. Tighten (toward Build Mode): "Shorten this to 3 bullet points suitable for a quick team update." 3. Deep dive: "Now drill into the top 5 drivers of gross margin change by product line." Short, directional follow-ups like "narrow to enterprise", "add charts", or "make this more executive-ready" work well. Because Concourse is transparent by design, you can always look at the reasoning and SQL, adjust logic, re-run queries, and converge on a version you trust. # Shape the output format Source: https://docs.concourse.co/concourse-101/prompting-best-practices/shape-the-output-format Tell Concourse how you plan to use the answer. * **Tone and density** – "Keep it under 5 bullets", "Explain it like I am new to the business", "Assume an expert finance audience". * **Structure** – "Use a table with columns: metric, current period, prior period, variance, variance %", "Group by department, then sort by variance descending". # Start with the job-to-be-done Source: https://docs.concourse.co/concourse-101/prompting-best-practices/start-with-the-job-to-be-done Before you type, think about what you would ask an analyst. Good starter frames: * "Show me ARR by customer and highlight the top 20 logos by change vs last month." * "Give me a first-pass monthly close summary for \." * "Prepare something I can paste into a Slack update for the exec team about \." You can always tighten or redirect on the next turn.