Automate · Quoted per workflow
Custom AI workflows with a human where it counts
- Published
- Published
- Last reviewed
- Reviewed
The short answer
A custom AI workflow uses a language model inside a process you already run: reading inbound email and routing it, summarizing calls, drafting replies for approval, extracting fields from documents, or scoring leads. Redenn scopes and builds each workflow to a fixed quote, with a human checkpoint wherever a wrong answer would cost you money or trust.
What changes
- Inbound email sorted and routed before anyone opens the inbox
- Long calls, threads and documents reduced to a summary with the next action
- Replies drafted in your voice and waiting for a one-click approval
- Data pulled out of PDFs and photos into your system without retyping
What is included
- Workflow scoping with a clear definition of what the AI decides and what a person decides
- Model selection and prompt design, using your own data for grounding
- Confidence thresholds and a review queue for uncertain cases
- Logging of every input and output for audit
- Integration with email, CRM, storage or your application
- Monthly review of accuracy and edge cases
How it works
- 01
Find the step
We look for one repetitive judgement in your process that a model can make well enough with a check.
- 02
Prototype
A small prototype runs on your real data so you can see where it is right and wrong.
- 03
Build with checkpoints
The workflow is built with thresholds, a review queue and logging.
- 04
Run and measure
Accuracy is reviewed monthly; the human checkpoint stays until the numbers justify moving it.
Who this is for
Businesses drowning in text: a trucking dispatcher's email, an immigration consultant's document piles, a law firm's intake forms, an accountant's client questions, a manufacturer's quote requests, a clinic's referrals. The AI handles the reading; people keep the decisions.
It is not for a process that a simple rule can handle. If 'route emails with the word invoice to accounts' does the job, that is a CA$149 Automation, not an AI workflow, and we will say so.
What we build
| Workflow | Input | AI does | Person does |
|---|---|---|---|
| Inbox triage | Inbound email | Classifies, extracts key facts, routes | Handles the routed item |
| Call summary | Call transcript | Summarizes, lists actions, updates CRM | Confirms actions |
| Reply drafting | Customer question | Drafts a reply from your policies | Edits and sends |
| Document extraction | PDF, scan or photo | Pulls fields into a record | Reviews low-confidence fields |
| Lead scoring | Form and chat data | Scores fit and urgency | Decides who to call first |
How we keep it honest
Every workflow is grounded in your own data rather than the model's general knowledge, carries a confidence threshold below which a person reviews the case, and logs every input and output. You can see what the model saw and what it decided, for every record.
Model choice is made per task and documented: which provider, where the data goes, and whether the provider's terms exclude your data from training. If a requirement cannot be met by a provider, we say so before building.
What it costs
Each workflow is quoted after a short prototype on your data, so the quote reflects real accuracy rather than a guess. Running costs are Redenn Dynamic hosting at CA$49 per month or Pro Managed at CA$99 per month, plus AI model usage billed at cost above the included tier.
Limitations and honest notes
Language models make mistakes, sometimes confidently. The human checkpoint is not optional for anything involving money, legal status, health or a customer's trust.
Personal information sent to a model provider is a disclosure under PIPEDA or provincial law. The scope names the provider and its data terms so you can decide.
Redenn does not build AI workflows that make final legal, medical or financial decisions, and does not promise time savings we have not measured on your own data.
Questions
Which AI models do you use?
The one that fits the task, chosen during the prototype and named in the scope along with the provider's data terms. You always know where your data goes.
Is our data used to train the model?
We use provider settings and agreements that exclude your data from training where the provider offers them, and we state the position in the scope. If a provider cannot commit, we choose another.
How do we know the AI is right?
Every output is logged, low-confidence cases go to a person, and accuracy is reviewed monthly against what your team corrected.
What does it cost to run?
Hosting from CA$49 per month plus model usage at cost. The prototype gives a realistic monthly usage estimate before you commit.
Related
Services
- Document AutomationQuotes, contracts, intake forms and reports generated from data you already hold, from CA$149 per document.
- Lead Follow-Up AutomationInstant reply, reminders and hand-off rules so no enquiry sits unanswered, from CA$149 one-time.
- AI ReceptionistAnswers the calls you miss, captures the need, books or routes, and messages you the summary, from CA$149/month.
- Custom SoftwareFixed-scope software for processes that off-the-shelf tools handle badly, with a build-or-buy checklist first.
- Business DashboardLeads, jobs, bookings, cash and search data from your existing tools on one screen, quoted per project.
Guides
Automate it. With custom ai workflows.
Configure a website in minutes, or ask for a written quote. Quotes arrive within two business days.