There is a clear and growing demand for personalised, bespoke workflows that are deterministic (predictable and rule-based), incorporating human input alongside integrations with third-party systems and services like Artificial Intelligence (AI) prompts, CRM updates, PDF generation, email communication, and other automated tasks. These workflows address unique business processes, reduce errors, ensure compliance, and enable scalable operations.

Businesses across sectors such as finance, healthcare, manufacturing, education, and retail are increasingly adopting bespoke workflows to handle complex, hybrid human-system interactions, where humans provide oversight or decisions, and systems facilitate seamless data flow and automation.

This need for bespoke, deterministic workflows stems from several factors:

  1. Customisation for unique processes: Off-the-shelf software often fail to capture proprietary or industry-specific workflows, leading to inefficiencies. Bespoke solutions on the other hand, allow tailoring to exact needs, such as integrating AI for predictive analysis while maintaining human approval loops for critical decisions.

  2. Determinism for reliability: In regulated industries, probabilistic AI (e.g., non-deterministic agents) can introduce risks like unpredictable outcomes. Deterministic workflows provide auditability, reproducibility, and consistency, making them essential for compliance, financial processes, and supply chain management.

  3. Hybrid human-system integration: Modern workflows blend human input (e.g., approvals, review, creative decisions) with systems for AI (e.g., ChatGPT prompts), CRM (e.g., Salesforce updates), sending email or SMS communication to clients, and other services. This enables input-output cycles that enhance productivity without fully automating away human roles.

  4. Evidence from trends: Recent discussions highlight a shift toward hybrid models combining deterministic structures with AI enhancements. For instance, businesses are using low-code platforms such as Kotive to run custom workflows that incorporate AI agents for flexibility but retain deterministic backbones for core operations.

Definition of a deterministic workflow: a predefined, rule-based sequence of steps in a business process where the same inputs always produce the exact same outputs and follow the identical path every time, with no randomness, variability, or reliance on probabilistic decisions.

AI’s probabilistic nature: it refers to the inherent uncertainty and randomness in AI models, where outputs are generated by sampling from probability distributions over possible responses. This means the same input can produce different (though often similar) outputs each time, even under identical conditions.

Example of a deterministic workflow in Financial Services

In the financial industry, a deterministic workflow for new client onboarding ensures regulatory compliance (e.g., AML, FICA, or equivalent standards) through a predictable, rule-based sequence that begins when a prospective client submits an initial application via a secure online form.

  • The prospective client submits required information and documentation (ID proof, address verification, source of funds, beneficial ownership details for entities, etc.).
  • Once uploaded, the workflow extracts and validates data using OCR if needed, then integrates with a third-party KYC/AML provider to perform automated identity verification, sanctions screening, and PEP checks.
  • Results feed into a fixed scoring logic that generates a scorecard summary report (e.g., risk tier: low/medium/high, with weighted scores based on predefined criteria like match confidence, red flags, and documentation completeness).
  • If the scorecard meets acceptance thresholds (or after mandatory review by the Key Individual for high-risk cases), the workflow consolidates all verified data, documents, report, and client details into a unified client profile created or updated in Kotive.

This workflow establishes a single source of truth for the client’s record, enabling seamless activation of accounts, ongoing monitoring triggers, and full audit trails while minimising manual errors and ensuring repeatable outcomes every time.

 
 

 

Example of a deterministic workflow in CPD Training

In the CPD (Continuing Professional Development) training industry, particularly in South Africa where providers offer accredited workshops, webinars, and courses for professionals needing points (e.g., ECSA, HPCSA, FPI, or similar bodies), a deterministic workflow streamlines customer registration for training events.

  • When a prospective attendee registers via an online form or event portal, providing their personal and professional details (name, contact info, profession, ID number if required), if they’ve agreed, the system automatically adds them to a targeted mailing list (e.g., via Mailchimp or integrated CRM) for event reminders, updates, and future offerings.
  • If the customer requests a quote (common for corporate-sponsored or bulk registrations), the workflow auto-generates a personalized PDF quote document, detailing the event name, date, fees, VAT, discounts, and terms, by using fixed templates and data merging.
  • It then emails it directly to the customer while cc’ing the finance department for visibility and tracking.
  • Upon acceptance, the customer clicks a secure payment link embedded in the quote email, redirecting them to a Payfast payment form (a popular South African gateway supporting cards, EFT, and mobile options) where they complete the online transaction.
  • On successful payment confirmation, the workflow auto-subscribes the customer to the online component of the training. They are provisioned login credentials, granted access to a custom e-learning platform, enrolled in relevant modules, and sent a confirmation email with access instructions.

This workflow ensures seamless progression from interest to participation with full auditability, reduced manual handling, and compliance with local payment and data regulations.

Example of a deterministic workflow in Tertiary Education

In educational institutions that enforce a strict no-AI policy for assessments, a deterministic workflow automatically screens every student submission to uphold academic integrity.

  • Upon upload (digital or scanned handwritten work), the system first converts the content to plain text using OCR APIs if necessary.
  • The extracted text is then analysed via a custom, detailed AI prompt, which outputs a report including an AI probability percentage (0–100), confidence level, key indicators of AI generation, and a recommendation.
  • Fixed institutional thresholds then determine the next step in the process. Submissions:
    • below 25% AI probability are auto-accepted and routed to lecturers for grading along with the score report,
    • between 25–60% are flagged for lecturer review, and
    • those exceeding 60% (or high-confidence flags) are auto-rejected with a notification to the student, preventing them from being graded until resolved.
  • Lecturers receive the original assessment, and full AI Probability Score Report as supporting evidence, ensuring consistent, auditable enforcement while maintaining human judgment in the final assessment.