Aide

Standpoint

Every brief has a standpoint. Here’s ours.

If you run an agenda pack through ChatGPT, Gemini, or Grok, you aren’t getting a view from nowhere. You get an answer shaped by internet training data, then reshaped by what the tool has learned about your personal preferences.

That is a standpoint — it just rarely admits it.

Not this framing

Biased vs. neutral

A false choice. Every useful brief comes from somewhere — including ones that sound balanced.

The real choice

Hidden standpoint vs. declared method

Aide also has a standpoint. We name it, bind it to the pack, and keep it inspectable.

The two hidden biases

Source bias

Public AI feeds on news, opinion, advocacy, and social media. That diet is not a representative sample of your local community. It carries implicit assumptions about rates, housing, and risk — and presents them with smooth confidence.

Mirror bias

Public chat products are built to be agreeable. If you lean toward caution, the answers become caution-shaped. Combine that with source bias, and you get a feedback loop: the internet’s blind spots merged with your own, handed back as certainty.

That is a governance risk. You need to stress-test a staff paper, not rehearse a pre-existing worldview.

Inspectable decision support

Aide uses modern AI models, but anchors them strictly to your agenda pack, not the internet average. Our standpoint is explicit:

Anchored strictly to the pack

Statements require page citations. Uncited fluff is blocked.

Consequence-focused

We highlight second- and third-order impacts — clearly labeling inferences so they aren’t smuggled in as facts.

Trap detection

Aide flags blank Te Tiriti fields, missing baselines, unpriced unit rates, deferred expenditure, and unmonitored financial exposure.

Gaps stay visible

Missing measures aren’t smoothed over with generic boilerplate; they are explicitly called out.

Hard line on the vote

Aide never recommends a vote, drafts political spin, or invents consensus.

General chat vs. Aide

Source lens
General chat AIWeb training + your chat history
AideDeclared decision-support framework
Optimization goal
General chat AIFluency, agreeableness, engagement
AideMaterial issues, risks, missing data
Missing information
General chat AIFills silence with generic web knowledge
AideLeaves gaps visible; stays quiet
Evidence standard
General chat AIText summary without strict sources
AidePage-cited to the council pack
Political stance
General chat AIEasily drifts into advice / coaching
AideFixed boundary: zero vote recommendations

Every brief carries a perspective.

When people ask, “Isn’t all AI biased?” the answer is yes.

The choice is whether that perspective is an undeclared mix you have to second-guess under pressure, or a transparent method you can inspect, verify, and argue with.

Aide built the second one.

Request access to Aide