Turbocharging fairer, faster, better with Tribepad Sidekick
Everything we do at Tribepad is driven by our mission to make recruitment fairer, faster, and better for everyone.
We’re devout believers in technology as a springboard for change. But we also refuse to hurtle headlong; to leap onto bandwagons that champion short-term gains at the cost of long-term progress.
That’s been our approach to AI. While competitors have leapt to capitalise on market hype with AI bolt-ons, we’ve taken the time to step back. To understand the full, long-term implications, consequences, challenges, and value of AI before developing.
The result is Tribepad Sidekick
Unlike competitor AI products, Tribepad Sidekick is a fully-integrated, in-house built AI engine that aligns with our mission and philosophy. We built Tribepad Sidekick on Amazon Bedrock, which uses the industry-leading Claude LLM.
Ethical
Many AI tools are a black-box, wreaking havoc on fairness. We built Tribepad Sidekick in-house, for total transparency and cast-iron data security.
Integrated
Tribepad Sidekick isn’t a bolt-on or plug-in. It’s a core engine powering fairer, faster, better recruitment from vacancy to hire.
Predictive
We tested Tribepad Sidekick against actual client data, to ensure the engine is aligned, fair, and effective at predicting positive real-life outcomes.
Collaborative
Tribepad Sidekick will never make decisions for you. It’ll free up time from tedious admin and surface better data, so you can make better decisions.
Evolving
Tribepad Sidekick isn’t a hodge-podge of sewn-together APIs. We own it – so we’re free to evolve. It’s a recruitment partner who gets better over time.
FAQs
What LLMs are used?
We use a large language model on AWS (Bedrock) called Claude (owned by Anthropic).
Can a Tribepad customer use their own LLM?
No, there are no current plans to allow this.
How is Sidekick trained?
We don’t train our own models, for part of Sidekick (skills and questionnaire matching as well as skills score) we use AWS Bedrock platform with anthropic.claude-3-haiku LLM – so this is trained by Anthropic and for the CV parsing, SWOT analysis and parts of the scoring, we use HRFlow, which is trained by their own teams. You can learn more here
Who has built Sidekick, has it been built by Tribepad or an external party?
Sidekick has been built by Tribepad in-house engineers working with external tools.
Does the AI comply with relevant regulations (e.g., GDPR)?
The data is processed in AWS Bedrock so is compliant
Are there safeguards to prevent data leakage or harmful outputs?
Sidekick is protected using Amazon Bedrock Guardrails. https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails.html
If we use Sidekick, will our personal data be transferred outside the EU?
No – everything stays in the eu-west-2 region of AWS
What types of encryption is used to protect information shared?
Sidekick will be subject to the same levels of web based SSL security as all other features of Tribepad,you can read more on this here. You can also read a summary of HRFlow’s approach to APIs here – any data stored in our database is subject to the normal security standards we already have in place which you can read about here.
What are the standards of data retention for using Sidekick, and how long is any data stored?
Tribepad only stores non-sensitive data via AWS. For details on data retention policy for HR flow, you can read more here.
Is our data segregated from other customer information who also utilise the AI functionality?
Yes, like your ATS, the data sent to HR Flow is linked to a specific customer account and this is segregated from all other Tribepad customers
Will our data be used to train Sidekick?
We will not use your data, The HrFlow AI learns only based on human feedback. Here is the paper that explains how their system is built and follows the 4 main principles of auditability (documentation, traceability of the predictions, explainability, accountability of our R&D team) https://blog.hrflow.ai/building-a-human-level-matching-algorithm-our-approach-for-ai-fairness/
Is data collected whilst using Sidekick then used to train AI models of other businesses?
No data is ever used to train models, we use models stored on AWS bedrock
How are Tribepad ensuring models that power Sidekick are kept up to data?
Sidekick, like all features and facets of Tribepad’s core ATS system are maintained regularly throughout the year, and LLMs will always be kept in line with the latest models from AWS
What data does the AI functionality access, process, or store?
We access job information, basic fields including job title and location as well as job advert data. This is then sent to our partner HrFlow for processing. Any personal data uploaded into the candidate profile will be used. CV, name, address, education, career etc. This is passed over to HRFlow for CV Parsing purposes, and also used in AWS Bedrock for Skills matching. Tribepad only stores non-sensitive data via AWS. For details on data retention policy for HR flow, you can read more here – https://developers.hrflow.ai/docs/gdpr
Are AI interactions logged and auditable?
Sidekick’s job description generator will log input data, output responses and user ID. For details on data retention and auditability, you can read more here.
What personal data will be processed if we use Sidekick?
Any personal data uploaded into the candidate profile will be used. CV, name, address, education, career etc. This is passed over to HRFlow for CV Parsing purposes, and also used in AWS Bedrock for Skills matching.
Do candidates have the option to choose what personal data is used by Sidekick?
No, processing of data is standard for each role and applicant.
Is it mandatory that customers using Sidekick add information to their Privacy Policy?
We recommend adding Sidekick information to our customers’ Privacy Policy pages and documents, though it is not a mandatory part of using the product.
Is there evidence to show that Sidekick works and results are valid?
Early results from the proof-of-concept shows that yes, it does. It has been tested with 3 existing Tribepad customers with the data showing that Sidekick has correctly predicted the top three candidates in 14 out of 15 test cases. Sidekick has also been robustly tested with a selection of customers prior to launch.
Will Sidekick automatically reject candidates?
This isn’t currently planned – Sidekick is intended to augment decision-making for recruiters rather than taking automated decisions on candidates.
How can customers using Sidekick audit the results of candidate assessment?
Sidekick candidate scoring is a recommendation only, and can be switched on or off. Customers are free at any time to review all applicants and compare their own judgements to those of Sidekick in ensuring they select the correct applicants for interview.
How does salary generator work technically?
Our LLM (Claude) uses predictive text based on the data it has been trained on – predominantly using job title along with some weights (the model’s learned parameters from training) and balances (combination of training objectives, behaviours, and system-level integration) to generate the recommended numbers
How does the job description generator work?
Our LLM (Claude) responds to pre-written prompts e.g. “Give me a job description for this job title, location, using this tone of voice XYZ and these company values ‘XYZ’.” The job description is then produced using pieces of things Claude can source from similar examples.
Will Sidekick features be sold separately or as a whole subscription?
Currently Sidekick is sold as a single product.