io4 Agents · Professional services firms
An agent that automates the preparation of recurring proposals and supports your teams in writing complex statements of work: needs scoping, section-by-section drafting and quality control. You keep control of the decisions and commitments.
30 minutes, no commitment. We start from your process and show you the one that works at our place.
The agent reads the transcript and finds the client file, the prices and comparable proposals.
It prepares the analysis, asks its questions and proposes a scope to price.
You clarify and validateYour template, your rates, your rules. The agent produces the document and flags the points to review.
You review, decide and authorize the planned actions. Sending to the client stays human.
You keep the decisionAutomate and support the writing
The level of automation depends on the mandate. Within a single statement of work, the agent can prepare the repetitive sections and support your team on the ones that require more thought.
Both modes can coexist in the same file. Validations also happen during preparation.
Your recurring proposals
Your proposals follow a known structure, with well-defined sources and rules. The agent takes on the assembly work up to the planned validations.
Your team focuses on the trade-offs and the approval.
Your complex SOWs
Several areas of expertise, requirements to track and validations to coordinate: the agent supports your team section by section, including for files of 100 to 120 pages.
Your team builds the proposal with a writing and review guide.
Review, in context
Each requirement is linked to its source and to the section that answers it. Your team tracks progress and reviews the flagged points before validating.
Illustration of the guided journey. The document, the controls and the interface are tailored to the mandate.
Point to checkReconcile the scope, the hours and the price before approval.
The starting point: accessible sources, explicit rules and a validation owner. For a complex SOW, you can start with one section or with the review, then broaden the support.
The proof starts with us
Same team. Same volume. Less time preparing each statement of work.
Results reported by io4 for its internal process: 12 proposals a week, prepared by a team of 3 salespeople. The gain on your own files is assessed in the pilot. For a complex SOW, writing, review and correction time are measured separately.
The agent checks the CRM and comparable proposals, prepares the scope, fills the template and checks the document. The preparer steps in for missing information, trade-offs and approvals.
The detailed case shows reconstructions of the process with a fictional client, without exposing any client data.
The expertise stays in your team.
The assembly work takes up less room.
An assistant to think and write
The agent prepares the questions to ask, surfaces the missing information and structures your answers. The salesperson and the experts clarify the commitments before writing.
A better-defined need gives a stronger base for the scope, the price and the delivery.
Quality control, concretely
Illustrative examples of checks to configure with your team. Each signal is reviewed and validated by the preparer.
| Point to flag | Proposed check |
|---|---|
| The scope covers 10 sites; the estimate covers 8. | Reconcile the scope, the hours and the price before approval. |
| The client asks for training; no deliverable addresses it. | Add the answer in the proposal or have an explicit exclusion validated. |
| A date depends on data the client must provide. | Specify the prerequisite, the expected date and the responsible person. |
| Two sections assign different responsibilities. | Have the roles arbitrated, then align the sections concerned. |
Your tools. Your way of working.
Microsoft 365 and your CRM remain your systems of record. The agent finds the context there to prepare your documents or support you in writing them, by your rules. Microsoft Copilot Studio, Claude or OpenAI: the choice depends on your environment and the work to be done.
Your Microsoft usage
We also build agents with Microsoft Copilot Studio, connected to your sources and processes through Power Platform connectors and flows.
An option to use the agent inside Teams or Microsoft 365. We validate with your IT team the licenses, access and actions needed for your use case.
The approach we use at io4
We use Claude for our own proposal process. An enterprise subscription and the authorized connections make it possible to prepare the work with users.
For execution integrated into your applications or triggered by an event, we assess a dedicated deployment and its consumption.
Already using ChatGPT?
A comparable approach can be built with OpenAI. We first assess your ChatGPT subscription, the available connections and the actions your company permits.
Does your team already have Pro accounts? We check what they allow for the intended use, and whether a Business or Enterprise environment is preferable for shared work and its administration.
With a user. The agent prepares the proposal in the chosen environment, asks its questions and waits for the planned validations. The available features depend on the plan, the connections and your organization's rules.
On an event or at a set time. An automation may require integration and usage-based billing that are separate. A ChatGPT subscription does not automatically cover calls to the OpenAI API.
Hosting and data. We document the flows, retention, processing region and applicable terms before deployment. For Claude via Microsoft Foundry, these depend on the offering and hosting model chosen; choosing an Azure region is not enough to guarantee that all processing stays there.
Copilot Studio can also fit. We assess it when the agent needs to fit into your Teams and Power Platform usage. Access to sources and write permissions are defined with your IT team.
From process to first use
We define the right balance between automation and guided writing: the sections the agent prepares, the questions it asks you, the checks to perform and the validations to obtain.
You bring an example. We look at the steps, the volumes and the tools, then show you the io4 process.
30 minutes · no commitmentWe define the sources, the rules, the prerequisites and the success criteria of the first process.
Scoping workshop · 2 weeksA pilot group tests the documents, the validations and the exceptions. We measure the time and the usage cost.
Pilot and adjustmentsTraining, documentation and handover come with the delivery. Follow-up is defined to fit your needs.
Production and evolutionNeed, scope, price and sending: the decisions stay human. The controls are tested on your cases, access is limited to the authorized scope and the actions to log are defined during scoping.
The time you win back must count
For your recurring proposals, adjust the assumptions below including review and validations. For a complex SOW, we estimate the gains per file with you: section preparation, review and corrections.
Estimated potential
1,150 hours of capacity recovered per year.
Provided all this capacity is turned into value. This calculation is not a guaranteed cash saving.
Budgets
io4 turns your sources, your templates and your way of preparing proposals into a workflow your team can use. Scoping defines the first scope, its prerequisites and its success criteria.
What your team receives.
Scoping questions, a writing plan and steps tailored to your proposal types. For complex files, tracking of requirements and section status.
Authorized sources, templates, terminology, rate card and rules connected to the agent within the agreed scope.
Review criteria, points to flag, owners and approvals defined with your experts.
A pilot on your cases, the agreed adjustments, training and documentation for use and operation.
To decide
Two workshops with your process owners to move from an intention to a plan ready to budget.
Fixed fee · 2 weeks
Credited toward the mandate if you proceed.
To put it in place
The four deliverables are configured for your first use, then tested with your proposal preparers. Automation, guided writing and controls are tuned to the agreed scope.
6 to 10 weeks · time and materials capped
Scope, document complexity and budget specified before signing.
Amounts are before tax, GST and QST extra. Subscriptions, model consumption and operating costs are separate. io4 commits to the means and deliverables defined in the statement of work.
Prepaid hour banks of 2,500 to 10,000 $ CA let you plan for adjustments, evolving connections and agent reviews. The budget and the interventions are agreed with you.
Once the proposals are in place, the approach can extend to mandate kickoffs, reports or project tracking. Each project has its own scope and estimate.
Explore the other automations and AI assistants →Illustrative scenario, to be validated in the pilot. An integration using the Claude API, for 48 short proposals a month. Long SOWs supported section by section are estimated separately. The amounts below are calculation assumptions, not a package or an io4 service rate.
| Poste | Assumption |
|---|---|
| Model consumptionUS$17.28, converted then rounded | ≈ 25 $ CA |
| Hosting and connectionsIllustrative allowance, depending on the architecture | 50 $ CA |
| Maintenance2 h / quarter × CA$190/h ÷ 3 months | ≈ 127 $ CA |
| Illustrative totalConsumption, hosting and maintenance above | ≈ 202 $ CA/mois |
Microsoft 365 and CRM licenses already in place: excluded from this example. No Claude or ChatGPT subscription is counted in this API scenario. Any other paid tool required would be added to the estimate.
Assumption per proposal, all steps and reviews combined: 60,000 input tokens and 12,000 output tokens, including reasoning billed where applicable. These volumes are not a guaranteed conversion from the number of pages; they are to be measured on your documents.
48 × [(60,000 ÷ 1,000,000 × US$3) + (12,000 ÷ 1,000,000 × US$15)] = US$17.28/month. Exchange assumption: US$1 = CA$1.40, i.e. CA$24.19, rounded to CA$25.
Reference pricing for Claude Sonnet 4.6: US$3 per million input tokens and US$15 per million output tokens, verified on September 18, 2026. Calculation without cache discount, paid web search, extra transcription or regional surcharge. See the Anthropic reference ↗
Before we talk
The tool, the access and the quality are discussed from your real process.
Yes. The support adapts to the complexity of the file, whether it's 15 pages or 120. The agent helps structure the plan, prepare each section, link requirements to their sources and track validations. Your experts write or validate the parts that require their judgment.
Quality control is based on criteria defined with you: requirements covered, missing sections, consistency of deliverables and estimates, terminology and appendices. The goal is guided progress and structured review. Time, cost and level of assistance are assessed on your documents in the pilot.
Not necessarily. We look at your subscription, the available features, your connections and your company's rules. A solution with OpenAI may fit; Claude or Copilot Studio are also assessed depending on the need. Any integration, usage and maintenance costs are specified separately.
The preparer answers the questions, validates the scope and the price, authorizes the planned writes and reviews the proposal before sending it. We define these steps with you. An interactive agent waits for your answers; an event-triggered automation respects the permissions set for its scope.
The rules aim to surface missing information and inconsistencies to the preparer. The pilot is there to test these situations on your cases. Quality is assessed on the documents produced, the corrections needed and the controls applied; a score given by the agent does not replace human review.
The first step is to identify the right sources, the reference versions and the access. If this work goes beyond the initial scope, it's priced before proceeding. The agent needs a usable document base.
Your systems remain the systems of record, but the model receives the information needed for processing. We document the recipients, the access, the retention, the data usage and the applicable regions for the chosen solution, before deploying it with your IT team.
An agent for your next proposal
Recurring proposal or a 120-page SOW: bring an example. In 30 minutes, we spot what can be automated and the steps where a writing assistant would help. We show you our process in action.
No commitment · Your process · An io4 demo