ANSWERS / AI AGENTS & AUTOMATION
Practical answers about choosing a workflow, scoping a pilot and keeping people in control.
FuturesAI builds AI agents and workflow automations for small teams. Our services focus on enquiries and draft replies, document and data workflows, and reports and action tracking. We are based in Newcastle upon Tyne and work with teams across the UK. Explore the services.
An automation follows defined steps. An AI agent can use a model to interpret information and choose among permitted actions. A predictable task may only need a conventional automation; an agent is useful when the task needs interpretation. Both need clear boundaries and a way to handle exceptions.
Choose a frequent, bounded task with accessible inputs, a clear expected result and someone who can check the output. For example, preparing an enquiry reply for approval is easier to evaluate than handing over an entire inbox. Avoid starting with a process whose rules are still changing or whose mistakes would be difficult to detect.
We assess available APIs, permissions, licensing and the information that must move between systems. These are examples of systems to explore, not a guarantee that every connection is available or included. See how we scope a project.
Our focused custom-service pilot costs £750 one-off. Implementation is separately scoped from £2,000, and optional workflow care starts from £350 per month. Software, hosting and AI usage are additional unless included in the quote. Your written quote confirms the total payable and any applicable VAT. CVSift pricing is agreed separately. Read the full prices and inclusions.
One bounded workflow, agreed success measures, one prototype, one review round and written findings. Production rollout is not included, and the pilot fee is not automatically credited against implementation. The target is one week after the scope, inputs and required access are ready; the schedule is confirmed in writing.
Compare the existing process and the pilot using representative examples: task volume, time spent, errors, exceptions and human review effort. Include the full implementation and running costs. We agree the measures before starting and do not promise a fixed saving. See the proposed CVSift evaluation measures.
Access, ownership, licences and handover are agreed in your written project contract. Ongoing care is optional. If you cancel before the next monthly renewal, care ends at the end of the paid period without a new care charge for the next period. Separately billed hosting, licences or AI usage may still be needed. Read the website and service information.
Approval steps and permitted actions are agreed for each project. The example on this website shows a human reviewing a draft before the next step. That illustration is not a live agent, and the appropriate controls depend on the actual workflow.
No. Start with a description and fictional or suitably redacted examples. Before using business data, we agree the necessary information, providers, access, processing locations and retention. Do not send real CVs, patient records or passwords in an initial enquiry. Read how website enquiries are handled.
CVSift is a candidate-review prototype, with pilot enquiries open. The published recording uses fictional candidates. It demonstrates phrase coverage and evidence review; it is not evidence of hiring accuracy, fairness or client results. Watch and read the CVSift walkthrough.
The prototype helps a reviewer find material to check against role criteria and record their own findings. Phrase matches are not hiring recommendations. A recruiter must check the source CVs and remain responsible for decisions. A real-data deployment would require a separate assessment of data handling, security and the intended hiring process.
Tell us what repeats, which tools you use and where the process gets stuck. Use fictional or redacted examples.
Discuss your workflow